Levetiracetam

Three of the four studies measured the effectiveness of levetiracetam, while the fourth study measured the comparative effectiveness of levetiracetam and phenobarbital.

Liu et al. (2020)³³ measured the impact of levetiracetam by randomizing treatment-naïve infants to either valproate alone (N=50) or valproate plus levetiracetam (N=50). Valproate dosing was 40-50 mg/kg/day, and levetiracetam dosing was 20-30 mg/kg/day. The average age at treatment initiation was two years, but seizure types were not reported. Authors also did not report whether any patients received concomitant medications during follow-up. At 12 weeks, authors reported better results for infants receiving valproate plus levetiracetam for all eight outcomes: seizure freedom (32% vs 22%), ≥75% reduction in seizures (72% vs 50%), ≥50% reduction in seizures (96% vs 70%), quality of life (Quality of Life in Epilepsy 31 [QOLI-31] scores, mean 84 vs mean 60, scale range 0-100 where higher scores are better), daily living ability (Barthel index scores, mean 86 vs mean 62, scale range 0-100 where higher scores are better), and three cognitive ability scales (see data in Appendix C).

One concern with Liu et al. (2020)³³ is that the authors used several outcome instruments not intended for young children (the Mini-Mental State Examination, the Weschler Memory Scale-Revised in China, QOLI-31, and the Barthel instrument). We contacted the authors asking if these instruments were modified for use in young children, but received no response. However, as we drew no conclusions about these outcomes, their inclusion does not influence our main findings.

Arzimanoglou et al. (2016)³² performed a pre/post study of 101 infants across Europe who had received levetiracetam (mean daily dose 46 mg/kg/day). The average age at treatment initiation was 6 months, 43% had focal impaired awareness seizures, 34% had focal to bilateral tonic-clinic seizures, and 25% had focal aware seizures (see other seizure types in Appendix C). At a mean of five months on levetiracetam, clinicians considered both seizure type and seizure frequency in rating each infant on a 1-7 scale where 1=marked worsening and 7=marked improvement. Improvement was marked in a third of infants (33%, 28/85), 26% (22/85) had moderate improvement, slight in 13% (11/85), and either no change or seizure worsening in the other 28% (24/85, see details in Appendix C). Clinicians also judged changes in psychomotor development using the same 1-7 scale. On this scale, improvement was marked in 19% (16/85), moderate in 13% (11/85), slight in 21% (18/85), and either no change or some worsening in the other 48% (40/85).

Arican et al. (2018)³⁴ performed a pre/post study of 92 treatment-naïve infants who had received levetiracetam (10-60 mg/kg/day, with 52% taking 30-40 mg/kg/day). The average age at treatment initiation was six months, 58% had focal seizures, and 42% had generalized seizures. At a median of 12 months, 66% of infants were free of seizures.

For comparative effectiveness, Grinspan et al. (2018)²⁹ retrospectively compared outcomes of 117 infants who received levetiracetam monotherapy to 38 infants who received phenobarbital monotherapy. The median target dose was 25 mg/kg/day for levetiracetam and 5 mg/kg/day for phenobarbital. The average age at treatment initiation was not reported, but all were under 1 year old. All had nonsyndromic epilepsy; seizure types were focal in 57%, generalized in 24%, and mixed/unclear in 19%. The only outcome reported was “freedom from monotherapy failure”, which was defined as no seizures during months 4-6 months after treatment initiation AND no second ASM other than pyridoxine prescribed during the six months after treatment initiation. The unadjusted rates were 40% for levetiracetam (47/117) and 16% (6/38) for phenobarbital (odds ratio 3.6, 95% confidence interval [CI] 1.5 to 10, favoring levetiracetam). Due to the non-randomized design, the authors conducted numerous additional analyses to control for selection bias, and all such analyses still favored levetiracetam over phenobarbital. The authors’ best estimate for the odds ratio was 4.2 (95% CI 1.1 to 16), which was based on a propensity score analysis.

Topiramate

Three studies addressed the effectiveness of topiramate, and one of the studies also addressed the comparative effectiveness of topiramate and carbamazepine.

Kim et al. (2009)³⁶ performed a non-randomized comparative study that enrolled treatment-naïve infants who had received either topiramate (N=41; 3 to 9 mg/kg/day) or carbamazepine (N=105; 5 to 30 mg/kg/day). The topiramate group averaged 10 months old at treatment initiation, and seizure types were generalized seizures in 71%, partial seizures in 20%, and unclassified in 10%. The carbamazepine group averaged 8.4 months old at treatment initiation, and seizure types were generalized seizures in 47%, partial seizures in 44%, and unclassified in 10%. The study did not report whether patients received concomitant medications during follow-up. Outcomes were reported an average of 30.7 months after treatment initiation. At six months, the rates of seizure freedom were 59% for topiramate and 55% for carbamazepine. The rates of ≥50% seizure reduction were 73% for topiramate and 63% for carbamazepine. Smaller reductions, no change, or aggravations occurred in 27% of those receiving topiramate and 37% of those receiving carbamazepine.

Kholin et al. (2014)³⁷ performed a pre/post study of topiramate that reported subgroup data on 58 infants who were age one year or younger at treatment initiation. For this subgroup, the study did not report mean age, median age, seizure types, or doses. For the overall population (N=722 including 636 who were older than 1 year), 62% were also taking additional ASM (specific medications not reported). Patients had been receiving topiramate for an average of about one year. Among 58 infants, 19% were seizure free, and 55% had ≥50% reduction in seizure frequency (timepoint not reported). The remaining 45% either had a smaller reduction, or an increase in seizures, or the appearance of new seizure types.

Grosso et al. (2005)³⁵ performed a pre/post study of topiramate recipients who were refractory to at least one ASM and reported subgroup data on 37 infants (median age 11 months at treatment initiation). Neither doses nor seizure types were specifically reported for these infants (although the mean dose for all 59 enrolled patients was 5.2 mg/kg/day). Specific concomitant medications were not reported for the 36 infants, but for the full population, 37% were receiving one ASM prior to starting topiramate, 41% were receiving two ASM, and 22% were receiving three ASM. At three months after starting topiramate, 8% of infants were seizure free, and 54% experienced a ≥50% reduction in seizures.

Lamotrigine

Only one included study reported effectiveness data for lamotrigine. Piña-Garza et al. (2008)^30,38 performed a withdrawal randomized trial of lamotrigine vs placebo in which all infants first received lamotrigine, and then only the lamotrigine responders (those whose seizure frequencies reduced by 40% or more) were randomized to either continue lamotrigine or receive a placebo substitution. However, as the placebo comparison period was at most eight weeks, the randomized phase of the study did not meet our inclusion criteria for effectiveness data. Instead, for effectiveness, we included the long-term pre/post data reported by the long-term open label follow-on publication of the study.³⁸ In Key Question 3, we included harms data for both the randomized portion and the open label portion of the trial.

The open label study enrolled 204 infants (mean age 15.9 months at treatment initiation) with partial seizures whose seizures had not been successfully controlled on at least one ASM. Of the 204 infants, 125 infants had already participated in the randomized portion of the trial and their parents opted for continued usage of lamotrigine. The maximum lamotrigine dosage was 5.1 mg/kg/day for those on either valproate or a non-enzyme-inducing ASM, or 15.6 mg/kg/day for those on enzyme-inducing ASM. The concomitant ASM was enzyme-inducing in 59%, not enzyme-inducing in 30%, and was valproate in 11%. Seizure types were partial only in 75%, both partial and generalized in 23%, and generalized only in 1%.

At 24+ weeks’ followup, 13% (26/204) of infants were seizure free, and 61% had ≥50% reduction in seizure frequency. The median seizure reduction (from a mean baseline of 21 seizures per week) was 74%.

Phenytoin

Only one included study reported effectiveness data for phenytoin. Sicca et al. (2000)³⁹ performed a pre/post study of 55 infants treated with oral phenytoin. Thirty-three first received phenytoin intravenously for status epilepticus and continued receiving oral phenytoin for seizure prophylaxis, whereas the other 22 infants were only treated with oral phenytoin for prophylaxis. The mean age was 7.4 months (of note, this mean represents the average for full N=82 who had received phenytoin orally only, or intravenously then orally, or intravenously only). Doses were not reported for the N=55 infants receiving oral phenytoin, but seizure types were generalized in 51% and partial in 49%. Concomitant treatments were used by 93% of infants (percentages not reported for specific medications).

At three months, the rate of seizure freedom was 4% (2/55) and the rate of ≥50% reduction in seizure frequency was 9% (5/55).

Vigabatrin

Only one included study reported effectiveness data for vigabatrin. Jackson et al. (2017)³¹ reported data on 103 infants (mean age 8 months at treatment initiation) who were treated with vigabatrin (median dose 93.8 mg/kg/day at last followup which after about one year of vigabatrin treatment). Concomitant treatments included levetiracetam in 35%, topiramate in 31.1%, phenobarbital in 25.2%, and several other ASM. About 90% experienced “epileptic spasms”, 15% had focal seizures, and 10% had generalized tonic seizures, and other seizure types were at most 5% prevalence among the enrolled infants.

At an average of one year on vigabatrin, 38% of infants were seizure free (33/88 with long-term followup data), 73% of infants had ≥50% reduction in seizure frequency, and the mean percentage reduction in seizures was 97% (interquartile range [IQR] 43% to 100%; baseline seizure frequency not reported).

Rufinamide

Only one included study reported effectiveness data for rufinamide. Tanritanir et al. (2021)⁴⁰ performed a pre/post study of 103 infants (median age 20 months at treatment initiation) who were treated with rufinamide (median dose was 42 mg/kg/day at last follow-up which was a median of 15 months). Concomitant treatments included levetiracetam in 69%, topiramate in 39%, clobazam in 33%, vigabatrin in 32%, clonazepam in 20%, phenobarbital in 17%, ketogenic diet in 16%, zonisamide in 14%, valproic acid in 10%, oxcarbazepine in 7%, steroid in 7%, lacosamide in 5%, lamotrigine in 5%, and other treatments in 12%. All patients had epilepsy, and seizure types were tonic in 75%, “epileptic spasms” in 64%, myoclonic in 43%, generalized tonic-clonic in 23%, focal onset in 22%, atonic in 14%, absence in 10%, and clonic in 5%.

At a median of 15 months of treatment, 19% (20/103) were seizure-free, and 50% (51/103) had experienced at least a 50% reduction in seizure frequency. The median % reduction in seizures was 54% (from ~167/month at baseline to 90/month at followup). Twenty-three percent (24/103) had discontinued rufinamide due to a lack of efficacy.

Stiripentol

Only one included study reported effectiveness data for stiripentol. Yamada et al. (2021)⁴¹ reported a subgroup analysis of 95 infants with Dravet Syndrome who had received stiripentol (age range 0-2 years at treatment initiation, average not reported). The dose for the 0-2 age group was not reported, but for the larger population of 376 patients, the median dose after one year was 32.5 mg/kg/day. Concomitant treatments were not reported specifically for the 0-2 subgroup, but for the full population, 99% were taking sodium valproate, 93% were taking clobazam, 41% were taking bromide, and 41% were taking topiramate. Seizure types were not reported, but all patients had Dravet Syndrome.

The only reported effectiveness outcome was the physician’s judgment of the degree of improvement. Specifically, physicians rated improvement on a 1-5 scale where 1=marked, 2=moderate, 3=mild, 4=no change, 5=worsened. This was based on seizure frequency, duration, intensity, and the ability to undertake activities of daily living. At two years, 54% (50/92) were rated as having either “marked” or “moderate” improvement.

Seizure Freedom Data

Figure 3 displays the seizure freedom rates reported by eight studies at various time points (Kholin et al. (2014)³⁷ is not shown because the study did not report the length of follow-up). The rates ranged widely (4%-66%), and it is unclear the extent to which any specific medication caused any patient to experience seizure freedom. This may be due to many factors including the lack of control groups, concomitant medications or other factors.

Figure 3

Seizure freedom rates after initiating medications. CBZ: Carbamazepine; LTG: Lamotrigine; LEV: Levetiracetam; PHT: Phenytoin; RUF: Rufinamide; TPM: Topiramate; VGB: Vigabatrin; VPA: Valproate. The vertical bars display 95% confidence intervals. When multiple (more...)

Ketogenic Diet

Two RCTs assessed the effectiveness of KD and MAD. In the first RCT, El-Rashidy et al. (2013)⁴² conducted a three-arm RCT comparing KD and MAD to a control group (no-dietary-change). All enrolled patients were under 36 months of age (n=40). The KD intervention was described as the classic 4:1 KD, administered via a liquid formula (Ketocal 4:1 milk from Danone, Nutricia). In addition to the dietary intervention, all maintained the same doses of ASM (valproic acid, carbamazepine and/or clonazepam) throughout the study period.

In the second RCT, Kim et al. (2015)⁴³ randomized patients to KD or MAD (see a discussion of the between-group comparison in the next section on KD vs MAD). Although the study enrolled patients ages 1 to 18 years, authors provided a subgroup of patients 1 to <2 years of age. The KD interventions followed the 4:1 lipid to nonlipid ratio with a nonfasting initiation protocol. While the MAD adhered to the Johns Hopkins protocol, the authors did not report if the KD used conformed to the protocol.

In the first pre/post study, Suo et al. (2012)⁴⁴ prospectively enrolled 317 consecutive patients with intractable epilepsy to receive KD, of whom 147 patients were 0-2 years old and met criteria for inclusion. Infants received the Johns Hopkins Hospital protocol, with a lipid-to-non-lipid ratio of 4:1. However, there was variation in how the diet was initiated across the study. Depending on the date of KD initiation, the patient’s diet may have been self-prepared via the KD Meal Planner, KetoCal ketogenic formula, or a KD Meal Planner supplemented by various liquid milk, cookies, and set meals. Follow-ups were conducted either at an outpatient clinic or by phone call.

Wu et al. (2015)⁴⁵ conducted a prospective pre/post study of KD in 87 children diagnosed with drug-resistant epilepsy. The authors did not clarify whether enrollment was selective or consecutive. The intervention used the Johns Hopkins protocol, with the addition of 24 to 48 hours pre-diet fast in the hospital. Of the 87 children in the study, only 40 (46%) met our inclusion criteria of under 36 months of age. Of the 40 infants, six were between the ages of 0 to 1 year, and 36 were between the ages of 1 to 3 years (36 months).

Kim et al. (2019)⁴⁶ presented a retrospective chart review of patients with medically intractable epilepsy who started the KD. The KD protocol was determined internally with a multidisciplinary team, and was not described as being based on specific established protocols. KD was initiated at a 1:1 fat to carbohydrate + protein ratio and steadily increased to 3:1, but not all patients maintained on the 3:1 ratio. Twenty patients maintained the 3:1 ratio, 13 progressed to 3.5:1, while 59 were on 4:1 ratios or higher. Roughly one-third of patients were on solid food, liquid food, or both solid and liquid foods each. All patients were under 3 years of age, and we only extracted data on the subgroup of patients without West syndrome.

Dressler et al. (2015)⁴⁷ evaluated the seizure relapse rate with a retrospective chart review of children treated with KD between March 1999 to April 2014. The intervention followed the Johns Hopkins protocol, without an initial fasting period. Of the 115 patients with complete clinical and follow-up data, 58 (50.4%) were under 1.5 years of age at the start of the KD. The outcomes the authors reported specifically for this subgroup were seizure freedom and “responders”, defined as a reduction in seizure frequency of ≥ 50%. Multiple durations of KD were also reported (3, 6, 12 months) and an additional 6 months after the KD ended.

Kang et al. (2005)⁴⁸ performed a retrospective review of young patients with uncontrolled epilepsy treated with KD. The protocol used was the Johns Hopkins protocol, but not all patients had initial fasting and fluid restriction. Of the 199 patients enrolled in the study, 49 (24.6%) were < 2 years old. Follow-up was conducted up to 12 months.

Liu et al. (2021)⁴⁹ enrolled 41 infants diagnosed with refractory epilepsy unresponsive to two or more anticonvulsants. Although the study’s primary goal was to examine biological and biochemical effects of the KD on infants, it did report seizure frequencies for infants treated with KD. Study authors did not specify what dietary protocol was used and some infants received different KD protocols. Infants in the KD group were further stratified into various age groups, but each individual subgroup contained too few patients to be analyzed independently according to protocol. All patients were followed for 12 months.

Seizure Freedom

Five studies (1 RCT, 4 pre/post studies) assessing KD reported the proportion of infants achieving seizure freedom. No studies reporting on seizure freedom compared KD to a control.

One RCT (Kim et al. 2015),⁴³ compared outcomes for infants receiving KD or MAD. For KD, at both 3 and 6 months, 53% (9/17) of infants were seizure-free. (It was unclear if the same nine patients were seizure free at both time points). More discussion about the direct comparison of KD and MAD is presented in a later section.

Seizure freedom rates across the four pre/post studies ranged from 11.6% to 53%. Suo et al. (2012)⁴⁴ reported seizure freedom rate of 11.6% (17/147). Notably, the study had high treatment attrition: only 33 of 147 children remained on the diet and reported on seizure freedom. Wu et al. (2015)⁴⁵ reported seizure freedom rates of 25% (10/40) at 2 months, and 33% (13/40) at six months.

Dressler et al. (2015)⁴⁷ was not clear and consistent in its outcome reporting for this subgroup, with ambiguous information presented in the text of the article and some conflicting information between the text and the figures. The article reported 20 of 58 patients (34%) seizure free three months into the KD. At both six months and 12 months, 19 of 58 patients (33%) were seizure free. The author also reported that 15 of 58 patients (26%) were seizure free six months after the cessation of the KD.

Kang et al. (2005)⁴⁸ reported 16 of 49 patients (32%) seizure free at three months, 18 of 49 patients (36%) seizure free at six months, and 13 of 49 patients (26%) seizure free at 12 months. It is unclear at each timepoint how many patients were still on the KD, and the denominator of 49 patients represents the total number of patients that had initiated the diet. At 12 months follow-up, only 13 of the original 49 patients (27%) in the < 2 years subgroup were still receiving the KD.

Figure 4 shows the seizure freedom proportion experienced at various time points for the five studies included here. Kim et al. (2015)⁴³ reported data for both KD and MAD, which are each displayed in the figure. Seizure freedom rates for KD ranged from 25%-53%.

Figure 4

Seizure freedom rates after initiating dietary interventions. KD – Ketogenic diet; MAD – Modified Atkins Diet. The vertical bars display 95% confidence intervals. When multiple studies reported the same follow-up times, we used horizontal (more...)

Modified Atkins Diet

The Modified Atkins Diet (MAD) is a less restrictive dietary alternative to the ketogenic diet, with a focus on low carbohydrate intake but no restrictions on proteins, dairy, or fat. Two RCTs assessed the MAD. However, only one study, El-Rashidy et al. (2013)⁴² compared MAD to a control group (ASM polytherapy arm). The MAD demonstrated statistically significant reduction in seizure frequency compared to control at 6 months, but not at three months. Specifically, at six months, seizure reduction was 28% (MAD) compared to 8% (control), p<0.001. Compared to control, the MAD also demonstrated a statistically significant improvement in seizure severity (measured as Chalfont score) at both three and six months.

Ketogenic Diet Versus Other Diets

Two studies directly compared the KD with the MAD. El-Rashidy et al. (2013)⁴² reported that KD yielded significantly greater reduced seizure frequency at three and six months compared to the MAD. Specifically, at six months, seizure reduction was 71% (KD) vs. 28% (MAD), <0.001. However, there was no statistically difference in seizure severity (measured by Chalfont score).

Kim et al. (2015)⁴³ reported KD demonstrated a statistically significantly higher rates of seizure freedom compared to MAD, with 9 of 17 achieving seizure freedom for KD and 4 of 40 for MAD at three months. This difference was not sustained; at six months the two groups demonstrated statistically non-significant difference. All other outcomes (i.e., >90% and >50% seizure reduction) also demonstrated no statistically significant difference between groups. The authors did find that at six months, the KD yielded lower mean and median percentage of baseline seizures than the MAD.

Hemispherectomy/Hemispherotomy

Twelve studies (reported in 13 articles) described children 1 to 36 months old undergoing hemispherectomy/hemispherotomy. Several studies described a single center’s experience with performing different surgical procedures of which a subgroup of infants met our inclusion criteria. For instance, one pre/post study by Cook et al.⁵⁸ described surgical outcomes for 55 infants with cortical dysplasia undergoing hemispherectomy/hemispherotomy over a 16 year period. Sixteen of these 55 infants had hemimegalencephaly (HME) and outcomes for this subgroup are reported in Jonas et al.⁵⁹ (Authors also reported children undergoing surgery for infarction/ischemia, or Rasmussen Encephalitis; however, these groups did not meet inclusion criteria due to age). Infants underwent slightly different procedures based on year of operation: from 1986-1997 (14 anatomical hemispherectomies), from 1990-1997 (15 functional hemispherectomies), and from 1997-2002, 26 hemispherotomies).

Seizure Freedom

Eight retrospective pre/post studies^{50–52,56–59,61} (reported in 9 articles^{50,51,56–62}) reported on seizure freedom over a follow up period of 6 months to mean 4.3 years after surgery. Rates are presented in Figure 6. Studies included a combined 188 infants. One study (Otsuki et al.)⁵⁰ did not report follow up interval for subgroup of included patients. Three studies^58,60,61 reported seizure freedom at 6 months. Cook et al. reported a subgroup of 55 infants with cortical dysplasia undergoing anatomical hemispherectomy or functional hemispherectomy/hemispherotomy depending on the year the surgery was performed. At 6 months, 80% were seizure free. (Jonas et al. described a subset of 16 infants with HME; at 6 months 15 of 16 (93%) of these infants were seizure free). Two additional studies reported 9 of 14 (64%)⁶¹ and 7 of 10 (70%)⁶⁰ of infants were seizure free at 6 months.

Overall, seizure free rates at 1 year or longer ranged from 7% to 76%. Six studies^{51,56–58,60,62} described outcomes at 1 year or longer: Cook et al. reported 75% of infants were seizure free; other small studies reported seizure free rates of 50% (n=10)⁶⁰ and 76%⁵⁶ (n=21)⁵⁶ at one year and 58% (n=12)⁵¹ at two years. A larger study, Roth et al.^52,62 reported at follow up median 4.3 years, 70% of patients (30 of 43) were seizure free. In contrast to these higher proportions, Pinto et al.⁵⁷ reported only 7% (1 of 15) infants were seizure free at follow up of at least 1 year after surgery.

We used stringent definition of seizure freedom which required studies using the Engel classification to report Engel Ia. However, if we had considered Engel I as seizure freedom, we note that seizure freedom from the Pinto et al. study would increase to 66% (10 of 15); also, 4 other studies^52–55 reporting rates of 55% to 81% (consistent with the range of seizure freedom rates we already identified) would have been included.

Although pre/post studies have significant limitations due to lack of control group, in these studies most infants underwent surgery for medically refractory epilepsy. Their pre-surgical status, therefore, suggests strongly that none of them would have experienced seizure freedom if they had not undergone surgery. Also, while accurately capturing seizure counts using retrospective data from charts (i.e., not captured in the context of the trial) was felt to be a key study limitation, we felt seizure freedom would be much less subject to recall bias or other types of bias. The precise rates of seizure freedom varied greatly among studies. Thus, we conclude that some infants with medically refractory epilepsy achieve seizure freedom after hemispherectomy, and we rated the strength of the evidence as Low.

Figure 6

Seizure freedom rates after hemispherectomy or hemispherotomy. The vertical bars display 95% confidence intervals. When multiple studies reported the same followup times, we used horizontal offsets to improve visibility. For studies only reporting a minimum (more...)

Other Resective Surgical Procedures

Eight retrospective pre/post studies reported outcomes for infants undergoing non-hemispheric procedures. Five studies reported on seizure freedom, and 7 studies on seizure frequency.

Seizure Freedom

Five pre/post studies including a combined 70 infants reported on seizure freedom. Specifically, infants underwent focal cortical resections^61–63 (n=3), frontal or temporal lobe resection⁵¹ (n=1),⁵² and posterior disconnection⁶⁴ (n=1). Rates of seizure freedom ranged from 40% (Sugimoto et al.) to 70% (Loddenkemper et al.). Figure 7 presents seizure freedom rates and follow up durations. Specifically, after focal resection studies reported seizure freedom of 70% (7/10)⁶¹, 56% (9/16)⁶², and 40% (4/10)⁶³, at median 6 months, 24 months, and mean 3.2 years, respectively. Kalbhenn et al.⁶⁴ reported 50% (5/10) of patients were seizure free after posterior disconnection surgery at 2 years after surgery. Reinholdson et al. reported 50% (12/24)⁵¹ seizure freedom for infants undergoing⁵² frontal or temporal lobe resection⁵¹⁵²⁵¹ at 2 years after surgery.

Figure 7

Seizure freedom rates after other resection. The vertical bars display 95% confidence intervals. When multiple studies reported the same followup times, we used horizontal offsets to improve visibility. For studies that only reported a minimum follow (more...)

Tumor Resection

We identified 1 study Gaggero et al. 2009⁶⁶ focused exclusively on infants with epilepsy due to brain tumors. Authors performed a retrospective chart review of infants < 3 years referred to a single center (G. Gaslini Children’s hospital, Genoa Italy) for primary supratentorial brain tumors. Tumor location was cortical in 16, non-cortical in 4. Histologic tumor types included WHO Grade I (n=5) Grade II (n=4), Grade III (n=7), and Grade IV (n=5). Twenty children had epilepsy as a clinical manifestation with focal (n=12), generalized (n=8) and history of convulsive status epilepticus (n=5). Pre-surgery, ASM use was as follows: 1 ASM (n=9), 2 ASM (n=7), and 3 ASM (n=4). Eight had received chemotherapy, and 3 had received radiotherapy. Mean time from tumor diagnosis to surgery was 0.86 (standard deviation [SD] 0.63) months.

Post-surgical follow-up ranged from 4 to 10 years (mean 7.6, SD (3.74). We report seizure freedom rates described in the study here; however, because study authors did not appear to define seizure freedom as Engel IA, data from this study were not included in analysis of seizure freedom rates).

At 1 year after surgery, 9 of 20 patients (45%) were seizure free (of which 5 were off ASM). Seven had >90% improvement in seizure frequency (Engel II), two had >50% but <90% improvement (Engel III), and 2 had no change (Engel IV). At 4 years after surgery, 11 patients were seizure free (7 off ASM), 5 had only rare seizures (Engel II), 2 had worthwhile improvement (Engel III), and 2 had no change (Engel IV). Malignancy grade (but not histological diagnosis) was associated with seizure outcomes: specifically, children with low grade tumors were more likely to have a good outcome (Engel I or Engel II), p<.001, t=2.84). Three patients died from tumor recurrence within 4 to 4.5 years (2 choroid plexus carcinomas, 1 glioblastoma multiforme). Of 17 patients with follow-up at 8 years post-operative, 9 were seizure free (Engel I), 4 (Engel II), 2 (Engel III), and 2 (Engel IV).

Other Outcomes (All Procedures)

Only 4 pre/post studies reported on developmental outcomes (developmental quotient [DQ], language or functional status). Two studies reported DQ after hemispherectomy. Loddenkemper et al.⁶¹ included 24 infants undergoing hemispherectomy or focal resection (median age at surgery was 14 months [3 to 34]). Infants were evaluated at median 12 months (3 to 34) preoperatively, and median 24 months (10 to 53) after surgery using the Bayley scale. Although the proportion of infants with developmental delay (defined as DQ <70) decreased after surgery, this change was not statistically significant (p=0.125). However, of note, authors excluded 26 of 50 consecutive infants (due to incomplete data or use of other neuropsychological tests), limiting the generalizability of these findings. Another study (Jonas et al.)⁵⁹ found that in 16 infants undergoing hemispherectomy for HME, the Vineland DQ increased by 9.1 (SD 16) at 24 months (compared to 6 months). The spoken language rank also increased from 0.33 (SD 0.5) to 1.4 (SD 1.8) after surgery.

One study (Lettori⁶⁰) included 10 infants meeting inclusion criteria and undergoing hemispherectomy. Before surgery functional status of 2/10 infants was dependent (and status could not be assessed for 8/10). However, after surgery, functional status improved (6 dependent, 3 semi-independent, 1 independent). Finally, 1 study (Sugimoto et al.⁶³) reported some infants had improvement in developmental delay after undergoing focal cortical resection (8/10 with delay pre-operatively, 6 infants with improved or good status after surgery). However, the study did not report how delay was assessed.

Harms of Levetiracetam

One nonrandomized comparison study³⁶ and two pre/post/studies^32,34 reported adverse event data for levetiracetam. Please see Key Question 1 for summaries of their enrolled patients and treatment details.

We first discuss the reported data on levetiracetam discontinuation due to adverse events including “serious” events, “severe” events, or events requiring dose modification. We then summarize any adverse events that occurred in 10% or more of patients who received levetiracetam. Appendix C provides all adverse event data (including those with rates <10%).

Arzimanoglou et al. (2016)³² administered levetiracetam at a mean daily dose of 46 mg/kg/day. At least one treatment-emergent adverse event led to study discontinuation in 7% of patients (7/101). The specific events were respiratory disorder (2 patients), respiratory distress and infantile spasms (two patients), irritability (one patient), lower respiratory tract infection, psychomotor retardation (one patient), and respiratory failure (one patient). Also, 12% of patients (12/101) had a “severe” treatment-related adverse event, but authors stated that none were considered related to levetiracetam by the clinician. “Serious” events occurred in 32% (32/101), but only two of them (both convulsions) were considered levetiracetam-related (authors did not define the difference between “severe” and “serious”). Furthermore, 10% (10/101) had an adverse event that required dose modification.

Arican et al. (2018)³⁴ reported that among 92 patients taking levetiracetam 10-60 mg/kg/day (52% on 30-40 mg/kg/day), no patients discontinued due to adverse events. The study made no statements about serious or severe events or any events requiring dose modification.

Liu et al. (2020)³³ reported an RCT comparing valproate alone (40-50 mg/kg/day) to valproate + levetiracetam (20-30 mg/kg/day). The study did not explicitly report whether any infants discontinued levetiracetam due to adverse events, whether any events were severe/serious, or whether any dose modifications were necessary. Effectiveness data were reported for all enrolled patients at 12 weeks, so likely there were no <12-week discontinuations due to adverse events.

Regarding non-severe events experienced by at least 10% of patients, the only pertinent events were bronchitis (10% as reported by Arzimanoglou et al. (2016)³²) and convulsion (10% as reported by Arzimanoglou et al. (2016)³²).

Overall, the only small signal from these data involve possible respiratory harms of levetiracetam (as per the study discontinuations in Arzimanoglou et al. (2016)³²). However, neither of the other two studies reported any respiratory events.

Harms of Topiramate

Four studies of topiramate reported data on harms: two RCTs,^68,70 one nonrandomized comparative study,³⁶ and one pre/post study.⁶⁷ Both RCTs compared different doses of topiramate: Novotny et al. (2010)^69,70 compared placebo to three different doses (5 or 15 or 25 mg/kg/day; 37-38 infants per group), while Manitpisitkul et al. (2013)⁶⁸ compared four different doses (3 or 5 or 15 or 25 mg/kg/day; 13-15 infants per group). Dose comparisons are particularly informative, because the medication-harm causal connection is stronger if there is a clear dose-response association. Conversely, the absence of a no dose-response association reduces the likelihood that the medication causes the harm. We first discuss events causing medication discontinuation, then severe/serious events and dose modifications, and finally, events experienced by 10% or more of infants.

Reported harms in three of the four studies are summarized in Table 10 (the fourth study only reported hypohydrosis, so we discuss its results separately). For discontinuation due to adverse events as well as serious/severe events, rates were low in all three studies, with no dose-response association. For less severe events (occurring in at least 10% of patients) the two RCTs found dose-response associations for loss of appetite and upper respiratory tract infection. For three other less severe events, one RCT found a dose-response association, but the other RCT did not. For seven other less severe events, neither RCT showed a dose-response association (coughing, diarrhea, fever, viral infection, somnolence, otitis media, and rhinitis). Kim et al. (2009)³⁶ also reported a rate of 17% (7/41) for psychomotor retardation

The pre/post study by Kim et al. (2010)⁶⁷ focused only on the risk of hypohydrosis with topiramate, and authors had performed a subgroup analysis of 81 infants age 1 year or less. The rate of hypohydrosis in these infants was 48% (39/81). The only other study to mention this type of event was Kim et al. (2009),³⁶ which reported a rate of “anhidrosis” of 2% (1/41) (far lower than the 48% reported by Kim et al. 2010).⁶⁷

Overall, these adverse event data suggest topiramate is not difficult to tolerate. We did find consistent evidence for two non-severe adverse events with topiramate: loss of appetite (7%-20% with risk increasing with dose) and upper respiratory tract infection (8%-38% with risk increasing with dose).

Table 10

Summary of harms of topiramate.

Harms of Lamotrigine

Piña-Garza et al. (2008)^30,38 studied lamotrigine and reported three sets of data: the initial open label phase (N=177) for at least five 5 weeks, the RCT phase comparing continued lamotrigine to replacement of lamotrigine with placebo for eight weeks (N=19 per group), and the long-term open label phase (at least 24 weeks in 92% of infants; N=204). All harms data from these sets appear in Appendix C. In this section, we discuss only the RCT phase and the long-term open label phase (due to its larger N than the initial open label phase).

During the eight-week randomized phase, none of the 38 patients had lamotrigine discontinued due to adverse events; this may be because the only infants who entered this phase had demonstrated tolerance during the initial five-week open label phase. In the long-term open label phase, 9% of patients (18/204) had lamotrigine discontinued due to adverse events. The authors reported specifics for 15 of 18 discontinuations: pneumonia (n = 4), complex partial seizures (n = 3), status epilepticus (n = 3), rash (n = 3), and fever (n = 2).

Regarding serious events, two events occurred in the randomized phase: one lamotrigine patient (5%, 1/19) had serious bronchitis, and one placebo patient (5%, 1/19) had status epilepticus. During the long-term open label phase with lamotrigine, 8% had pneumonia (16/204), 6% (12/204) had status epilepticus, 6% (12/204) had complex partial seizures, 4% (12/204) had fever, 3% (6/204) had convulsion, 3% (6/204) had dehydration, and 3% (12/204) had gastroenteritis.

For non-serious events, the RCT phase had five events in which one of the two groups had rates of 10% of more (cough, nasal congestion, upper respiratory tract infection, fever, and teething). The long-term open label phase had 15 events with rates of 10% of more: fever, upper respiratory tract infection, ear infection, cough, vomiting, otitis media, irritability, constipation, nasopharyngitis, teething, rash, bronchitis, pneumonia, status epilepticus, upper respiratory tract congestion. See Appendix C for the rates of these events.

Overall, the most common adverse events reported with lamotrigine are fever (45%), upper respiratory tract infection (28%), and ear infection (22%). Left untreated, some of these events may later be associated with pneumonia, as 8% had lamotrigine discontinued for that reason.

Harms of Phenytoin

The pre/post study by Sicca et al. (2000)³⁹ enrolled 55 infants who had received oral phenytoin for seizure prophylaxis. The study reported neither phenytoin discontinuation due to adverse events nor any “serious” or “severe” harms of phenytoin. Of less severe events, only four had rates of 10% or more:

• Drowsiness: 22% (12/55)
• Gingival hyperplasia: 15% (8/55)
• Sleep troubles: 15% (8/55)
• Hyperactivity: 11% (6/55)

Although drowsiness is fairly common (22%), in general, these data are too sparse to draw clear conclusions regarding the adverse effects of oral phenytoin when administered to infants.

Harms of Vigabatrin

Jackson et al. (2017)³¹ was a pre/post study that administered vigabatrin to 103 infants. The authors reported a rate of vigabatrin discontinuation-due-to-adverse effects of 9% (9/103), with specific reasons being vision abnormality (n=5), fatigue (n=1), fatigue and anorexia (n=1), “possible vigabatrin toxicity” (n=1), and anemia (n=1). The study did not report rates of “severe” or “serious” adverse events.

Authors reported rates of adverse events specifically for the 71 infants who discontinued vigabatrin for any reason (e.g., the reason was successful control of seizures in 31 infants, and unsatisfactory therapeutic effect in 23 infants). None of the reported events reached 10% or more (see all data in Appendix C). Authors also specifically conducted eye exams before vigabatrin (N=49), during vigabatrin (n=62), and after vigabatrin (n=49) to evaluate for vision abnormalities possibly caused by vigabatrin. The authors reported these three time periods separately, but unfortunately did not report the number of infants whose vision status had changed between measurements (which would have more directly addressed the influence of vigabatrin).

Prior to vigabatrin administration, 69% (34/49) had vision abnormalities, which authors attributed to tuberous sclerosis complex, refractive errors, and prior medication. The study did not report which baseline medications were likely responsible for the pre-vigabatrin abnormal eye exam results. During vigabatrin, 81% (50/62) had at least one abnormal exam. After vigabatrin, 63% (31/49) had vision abnormalities.

Overall, some evidence suggests that vigabatrin may cause temporary vision abnormalities, but only a single pre/post study has addressed the issue.

Harms of Rufinamide

Tanritanir et al. (2021)⁴⁰ was a pre/post study that administered rufinamide to 103 infants. Fifteen (15%) discontinued rufinamide due to adverse effects (three of these were solely due to adverse effects, and the other 12 were due to both adverse effects and lack of efficacy). Authors did not mention whether any events were serious or severe. Rates of 10% or higher were observed for somnolence (12% or 12/103) and irritability (10% or 10/103).

Harms of Stiripentol

Yamada et al. (2021)⁴¹ was a pre/post study that administered stiripentol to 95 infants age 0-2 years. Authors did not report the specific types of adverse reactions in these patients, but did report that 61% (58/95) had at least one adverse drug reaction. Physicians in charge determined that all 58 had a causal relationship to stiripentol (either “clearly” or “probably” or possibly”). One patient died due to liver damage, which physicians deemed a “possible” relationship to stiripentol, but the patient was also taking both valproate and clobazam.

Harms of Ketogenic Diet

Of 8 included studies of ketogenic diet, 4 reported data pertaining to harms for infants age 1-36 months.^42,46,47,49 Three^43,44,48 of the other four reported data on adverse events, but data were not specific to age 1-36 months, and the fourth study⁴⁵ did not investigate whether adverse events occurred.

Two of four studies reported the rate of withdrawal due to side effects or diet intolerance. Kim et al. 2019⁴⁶ found a rate of 2% (2/109; one behavioral food refusal and one persistent acidosis), and El-Rashidy et al. (2013)⁴² found a rate of 20% (2/10; both diet intolerance).

For other reported adverse effects:

• Kim et al. (2019)⁴⁶ reported 33% of patients had decreased HC0₃ level (36/109), 32% constipation (35/109), , and 20% vomiting/reflux (22/109); three other events were experienced by <10% of patients.
• Dressler et al. (2015)⁴⁷ reported that 50% (29/58) of patients experienced “side effects”, but authors did not report specifics. Also, for 28% (16/58) there was “difficulty introducing solid foods”.
• Liu et al. (2021)⁴⁹ reported z scores of body mass index (BMI)-for-age at 3-12 months after KD initiation. For the 41 patients, the baseline mean was 0.49, which changed to 0.06 at 3 months, 0.16 at 6 months, and 0.35 at 1 year; none of these changes were statistically significant. Authors also reported statistically non-significant changes over time in the percentage of patients who were underweight (defined as weight-for-age z score <-2), stunting (defined as height-for-age z score <-2), and wasting (defined as BMI-for-age z score <-2). However, the percentage of patients who were overweight/obese decreased (statistically significantly) from 17% at baseline to 2% at one year after KD initiation.
• El-Rashidy et al. (2013)⁴² reported low rates of vomiting, constipation, diarrhea, and dysphagia (0%, 20%, 10% and 10%, respectively).

Harms of Modified Atkins Diet

One study reported harms data for Modified Atkins diet. Specifically, El-Rashidy et al. (2013)⁴² reported that 13% (2/15) of patients dropped out due to diet intolerance and had experienced “significant” weight loss. Also, El-Rashidy et al. (2013)⁴² reported 27% vomiting (4/15), 13% constipation (2/15), 13% diarrhea (2/15), and 20% dysphagia (3/15).

Harms of Hemispherectomy/Hemispherotomy

Eleven pre/post studies^{50–55,57,58,60,62,71} reported harms after hemispherectomy/hemispherotomy including mortality, hydrocephalus, infection, and other adverse events. One study (Roth et al.⁶²) included data from 19 centers with surgical procedures performed from 1999 to 2020. Seven patients cared for at 3 of 19 centers (University of California at Los Angeles, Cleveland Clinic, and Great Ormond Street Hospital) may have also been previously reported in other studies included in this report.^58,61,71 (author correspondence). A second study, Iwasaki et al. (2021)⁷² described harms for 75 infants, of which 9 hemispherectomy patients had previously been described in Otsuki et al. (2013)⁵⁰ (author correspondence). Further details are provided in Appendix C.

Mortality

Nine pre/post studies reported on mortality after hemispherectomy/hemispherotomy (see Figure 8). Studies described mortality after anatomical hemispherectomy (n=3), functional hemispherectomy or hemispherotomy (n=8), or across multiple procedures (lesionectomy, cortical resection, and hemispherectomy/hemispherotomy, 1 study). Only 3 retrospective chart reviews reported data for infants specifically undergoing anatomical hemispherectomy: Cook et al.⁵⁸ reported that of 14 infants (operated on from 1986 to 1997 at University of California Los Angeles) there was 1 intra-operative death, an 8 month old infant with HME. Dunkley et al.⁷¹ reported no deaths for either of the 2 infants undergoing anatomical hemispherectomy over a 10 year period (year not reported). Roth et al.⁶² reported no deaths for 10 infants undergoing anatomical hemispherectomy. Given the sparse data, we deemed the evidence insufficient to permit conclusions.

Eight studies described surgical mortality for a combined 196 infants undergoing functional hemispherectomy/hemispherotomy. Seven of 8 studies (including 180 infants)^{50,52,55,58,62,71,72} reported that there were no deaths; Kumar et al.⁵⁴ reported a single death of 16 hemispherotomies. The death occurred in an infant with epidermal nevus syndrome, right HME, and multiple other congenital abnormalities who developed refractory seizures after surgery and care was withdrawn.

Finally, Steinbok et al.⁵³ reported a single death across 116 infants undergoing 151 surgical procedures which were either a hemispherectomy/hemispherotomy, lesionectomy, or cortical resections. The intraoperative death occurred in a 3.9-month child with tuberous sclerosis undergoing attempted resection of intraventricular and extraventricular lesions.

Although limited, evidence from these studies suggests peri-operative mortality after functional hemispherectomy or hemispherotomy is uncommon. However, these studies were primarily single center retrospective chart reviews including heterogenous infants (with many different seizure etiologies), and studies often failed to specify what proportion of infants did not meet criteria for inclusion based on missing data. Furthermore, it is possible that centers with higher mortality rates might choose not to publish their data. However, despite these limitations of the evidence base, we concluded that surgical mortality after functional hemispherectomy/hemispherotomy is rare (SOE: Low).

Figure 8

Surgical mortality for hemispherectomy/hemispherotomy. *Nine infants undergoing hemispherectomy/hemispherotomy included in Iwasaki et al. 2021 (denoted Iwasaki[b]) were already described in Otsuki et al. **Roth et al. included data from 19 centers; 7 (more...)

Hydrocephalus and Other Adverse Events

Twelve studies reported other adverse events (AEs) for infants undergoing hemispherectomy/hemispherotomy (see Table 11).

For anatomical hemispherectomy, studies reported that development of hydrocephalus/ventriculoperitoneal shunt placement was common (see Figure 9). Three studies (combined n=19 surgeries) reported this AE: Dunkley et al. reported that both infants (2 of 2) required a VP shunt at 12 months after surgery. Similarly, 7 of 10 infants undergoing anatomical hemispherectomy in another study (Pinto et al.⁵⁷) required ventriculoperitoneal shunt placement. A third study (Lettori et al.⁶⁰) reported 3 of 7 infants undergoing anatomical hemispherectomy or hemidecortication developed hydrocephalus (follow-up interval not reported).

For functional hemispherectomy or hemispherotomy, 9 studies (combined n=196, plus infants from 1 additional study only reporting a percentage⁵² and 1 study with an unclear denominator⁵³) reported on this AE. Generally, studies reported lower rates of hydrocephalus/shunt placement compared to anatomical hemispherectomy. 1 of 9 studies reported no infants (0 of 10) developed hydrocephalus.⁵⁵ Another study reported 4 infants undergoing functional hemispherectomy developed hydrocephalus (at least 22 infants underwent functional hemispherotomy, but the total number of infants undergoing this procedure was unclear).⁵³ The remaining 7 studies reported rates of 8.3% (1 of 12),⁵¹ 11% (3 of 27),⁷¹ 16% (Kadish et al.⁵²), 20% (1 of 5),⁵⁷ 22% (6 of 27),⁷² 25% (4 of 16),⁵⁴ and 33% (1 of 3).⁶⁰

Finally, Roth et al.⁶² reported hydrocephalus in 25% (11 of 44) infants undergoing either anatomical hemispherectomy or functional hemispherectomy/hemispherotomy. Notably, only 1 study⁷¹ reported when hydrocephalus occurred. Another study reported a time range (months).⁵³ Four studies^52,55,55,60 did not report when hydrocephalus occurred, and remaining studies provided a timepoint at which other outcomes were measured (eg. >1 year after surgery), but no other information regarding timing of hydrocephalus.

Other AEs reported for anatomical hemispherectomy include:

• Central nervous system (CNS) infection (2/14)⁵⁸
• Deep infection (1/7)⁶⁰
• Superficial infection (1/7)⁶⁰
• Cranial nerve III palsy (1/14)⁵⁸
• Inappropriate antidiuretic hormone (1/14)⁵⁸
• Subdural fluid collection (1/7)⁶⁰
• Cerebrospinal fluid leakage (1/7)⁶⁰
• Transient fever (2/7)⁶⁰

Other AEs reported for functional hemispherectomy/hemispherotomy include:

• CNS infection (2/41)⁵⁸
• Superficial infection (1/3)⁶⁰
• Dural adhesions requiring “late reoperation” (1/41)⁵⁸
• Acute post-surgical seizures (23%)⁵²
• Epidural hemorrhage requiring surgical revision (1/22)⁵²
• Excessive bleeding (1/16)⁵⁴
• Pituitary failure due to thalamic lesion(1/22)⁵²
• Inadvertent extubation (1/16)⁵⁴
• Intraoperative disseminated intravascular coagulation (1/37)⁶²
• Cyst formation requiring surgical intervention (2/27)⁷²
• Cerebral salt wasting syndrome (2/27)⁷²
• Diabetes insipidus (3/27)⁷²
• Sinus thrombosis (resulting from diabetes insipidus) (2/27)⁷²
• Asymptomatic hemorrhagic infarction (1/27)⁷²

Finally, two studies reported combined AEs for multiple procedure types. One study (Steinbok et al⁵³) reported AEs for procedures including anatomical hemispherectomy, hemidecortication, functional hemispherectomies, and peri-insular hemispherotomies: 31 of 40 infants required blood transfusion, and 13 infants developed perioperative aseptic meningitis which was successfully managed with steroids.

A second study (Roth et al.⁶²) reported AEs for a total 69 procedures which included hemispheric and non-hemispheric surgeries. However, with the exception of one AE (disseminated intravascular coagulation), remaining AEs were not reported by procedure.

Table 11

Surgical studies reporting harms.

Figure 9

Rates of hydrocephalus after hemispherectomy/hemispherotomy. *Roth et al. included data from 19 centers; for hydrocephalus, 1 patient may have been previously described in Dunkley et al. Data from Kadish et al, is not shown here as authors only reported (more...)

Harms of Other Resective Surgical Procedures