We sought to address three Key Questions:

Key Question 1. What is the effectiveness and comparative effectiveness of pharmacologic treatments for infantile epilepsies (infants age 1 month to <36 months)?

Key Question 2. What is the effectiveness and comparative effectiveness of non-pharmacologic treatments for infantile epilepsies (e.g., dietary therapies, surgery, neuromodulation, gene therapy), including comparisons to other non-pharmacologic and/or pharmacologic therapies?

Key Question 3. What are the harms or comparative harms of treatments for infantile epilepsies?

This review also addresses two Contextual Questions:

CQ1. What are the parental preferences for treatment options for infantile epilepsies?

CQ2. What are the harms or comparative harms of not treating infantile epilepsies?

We address these two Contextual questions in relevant results sections, referring to evidence discovered during the review process.

Findings in Relation to What Is Known

Overall, as anticipated, the evidence base for management of infantile epilepsies was sparse. We identified only six RCTs, two non-randomized comparative studies, and 33 pre/post studies (mostly retrospective). In fact, for surgical interventions, despite only requiring studies to report outcome for 10 infants per procedure, we only identified 18 studies, and all 18 were retrospective pre/post studies. Given the difficult treatment decisions and high stakes faced by clinicians, families, and caregivers of infants with epilepsy, this limited evidence base represents an important evidence gap.

Effectiveness: Pharmacologic Interventions

Most infants with epilepsy receive a trial of pharmacologic interventions before other interventions including dietary or surgical treatments. However, our review found limited evidence for pharmacologic interventions (KQ1). Although we included studies assessing 10 drugs (levetiracetam, topiramate, lamotrigine, phenytoin, vigabatrin, valproate, phenobarbital, carbamazepine, rufinamide, stiripentol) evidence was only sufficient to demonstrate effectiveness for a single drug, levetiracetam (SOE: Low).

For other medications, limitations of the evidence included a lack of control groups, concomitant medications, insufficient follow-up time, only a single study, and/or unreported critical outcomes. Regarding comparative effectiveness, freedom from monotherapy failure appears to be more likely with levetiracetam than phenobarbital (one study). Another study compared topiramate and carbamazepine, but its data were inconclusive. None of the studies measured the effectiveness of oxcarbazepine, even though it is one of the more commonly prescribed medications for age 0-36 months.

Effectiveness: Dietary Interventions

Regarding dietary interventions (KQ2a), we found that for some infants, the KD is effective for producing seizure freedom and reducing seizure frequency. The MAD also may reduces seizure frequency. Furthermore, the KD was more likely to reduce seizure frequency than the MAD (SOE: Low). Most studies of dietary interventions lacked control groups, were relatively small (e.g., pre/post studies ranged from 40 to 147 infants), and the two RCTs each enrolled less than 40 infants and had considerable inter-study variability in the specific foods and ratios indicated for a KD. Dietary therapy is often only considered for infants in whom multiple ASM have been unsuccessful. In two surveys, parents of children on the KD or about initiate the diet rated fewer seizures as their top priority (65-69%); however, need for fewer drugs and improved cognition were their next priorities. Unfortunately, included studies failed to report on either of these outcomes.

Effectiveness: Surgical Interventions

For hemispherectomy/hemispherotomy, seizure freedom rates across seven studies ranged from 7% to 76%, and six of seven studies reported rates higher than 50%. The lower rate of 7% reported by Pinto et al. may have been related to epilepsy etiology: over half (8/15) included infants had hemimegalencephaly (HME), traditionally considered to be associated with worse outcomes. However, another study including infants with HME undergoing hemispherectomy/hemispherotomy reported a much higher rate of seizure freedom, 93% (15/16), at six months. Both studies included infants undergoing both anatomic hemispherectomy or functional hemispherectomy / hemispherotomy. In addition to the range of seizure rates, the heterogeneity of seizure etiologies, procedures and ancillary treatments, precluded conclusions regarding a quantitative estimate. Thus, we concluded that surgery may cause seizure freedom in some infants (SOE Low). These robust rates of seizure freedom reflect the fact that unlike ASMs, when successful, surgical interventions address the underlying cause of seizures.

We encountered similar variability across estimates of seizure frequency reduction. Although all studies found reductions in seizure frequency after surgery, the proportion of infants achieving one widely used definition of a favorable outcome (Engel I or II, ILAE I to IV, or >50% seizure reduction) ranged from 67% to 100% with most studies reporting outcomes at least one year. Due to this variability, along with important study limitations, evidence was insufficient to draw a conclusion regarding seizure frequency.

For infants undergoing other resections (frontal or temporal lobe resection, intralobar or multilobar resections, or posterior disconnection), rates of seizure freedom ranged from 40% to 70% over with follow up intervals from median 6 months to average 5.2 years (5 pre/post studies). Given study limitations and range of estimates, we concluded that other resections may cause seizure freedom in some infants (SOE: Low). Seizure reduction rates for these procedures was also variable, with 50% to 90% of infants achieving a favorable outcomes across variable endpoints (5 pre/post studies). Thus, evidence was insufficient to draw a conclusion regarding seizure reduction.

Parents may have particular concerns when considering surgical interventions. When asked about goals and priorities in considering epilepsy surgery for their child, parents of older children with epilepsy (age >36 months) rated seizure freedom as highest priority, followed by reduced medication, improved cognition, and greater independence. One small qualitative study found that parents choosing surgery saw this choice as last option, but the only real chance at a normal life. Although we only identified retrospective pre/post studies of surgical interventions, these studies do find that some infants achieve seizure freedom.

Harms

For the harms of treatment (KQ3), detailed reporting only exists for studies of pharmacologic treatments. We concluded that for three medications (levetiracetam, topiramate, and lamotrigine), adverse effects are rarely severe enough to warrant discontinuation of the medication. Specifically for topiramate, we did find consistent evidence of dose-response effects for two non-severe adverse effects: loss of appetite and upper respiratory tract infection. Although parents worry about both short and long term adverse effects from drugs, included studies only reported short-term harms and rarely measured neurodevelopmental outcomes. For diets, adverse effects were generally not reported.

Harms of dietary interventions were not well-reported, and evidence was insufficient to permit conclusions. Some parents may be concerned about dietary intolerance or the difficulty in maintaining special diets. We saw some suggestions of these problems in the evidence we reviewed, but not enough to draw clear conclusions.

We note that the long-term potential harms of pharmacological and dietary treatments remain unclear, since few studies followed patients for longer than one year. As many parents are understandably anxious to know about these long-term harms, particularly regarding neurocognitive development, long-term studies are particularly important for future work.

For surgical interventions, although studies reporting surgical mortality for functional hemispherectomy/hemispherotomy and multilobar, lobar, or focal resections reported rare rates (SOE: low), adverse event reporting across studies was inconsistent, and evidence did not permit estimates of the mortality risk. Inconsistent reporting across studies also precluded any quantitative estimates of rates of post-surgical adverse effects.

Regarding post-operative hydrocephalus, data were too sparse and patients too heterogenous to draw conclusions regarding development of hydrocephalus after hemispherectomy or hemispherotomy. Limited available data was consistent with other work suggesting hydrocephalus may be more common after anatomic hemispherectomy (compared to functional hemispherectomy/hemispherotomy). Notably, nearly all studies reporting on hydrocephalus failed to report when it occurred. However, a study in older children found that post-operative hydrocephalus can occur up to 8.5 years after surgery. Thus, rates of hydrocephalus may depend heavily on length of follow-up.

Implications for Clinical Practice

This systematic review confirms the lack of high-quality evidence to support treatments of infantile epilepsies. For many interventions currently in use, particularly pharmacologic treatments, we identified no studies assessing efficacy in these age groups. Nevertheless, our findings do support the use of levetiracetam, KD, MAD, and surgical procedures such as hemispherectomy, hemispherotomy, and other resections as effective in some infants. For studies of dietary therapies and surgical treatments, nearly all infants enrolled had intractable epilepsies, reflecting the use of these treatments only after drug therapy has failed.

Implications for Research and Health Policy

The substantial evidence gaps we identified have important implications for future research. As noted above, for many medications in current use, no studies in this age group exist. Although medications in other clinical areas are often evaluated using placebo-controlled RCTs, we acknowledge some important challenges to performing such trials in this population. Given fears regarding the potential for brain injury with prolonged seizures, parents may be understandably reluctant to risk their child receiving placebo. Such considerations have led some clinicians to consider a placebo arm unethical. Creative approaches to trial design, such the withdrawal RCT included for Key Question 1, may be necessary to address these issues. However, without randomization, inferring efficacy for drugs, dietary, and surgical therapies is likely to remain challenging given the number of concomitant therapies patients receive.

A key challenge we faced was the lack of full data reported separately for this age group. Many studies that likely included infants in this age range failed to report data for this subgroup separately, instead reporting outcomes for "pediatric" patients encompassing newborns to 18-year-olds. Interviews with key stakeholders suggested that neonates (age <1 month) also represent a clinically distinct group of patients. Thus, future studies should consider reporting data for these three subgroups (neonates, infants, older children) separately, to facilitate use of these data. In addition, future studies should also report baseline seizure frequency, seizure severity, and prior treatments.

Evidence for surgical interventions was particularly weak, with no studies assessing different approaches and inconsistent reporting of many clinical variables and key information such as follow up duration which are critical for assessing outcomes and adverse effects. Performing a controlled trial for surgical interventions presents obvious challenges, both ethical and pragmatic. Researchers could directly compare a surgical intervention to another treatment, such as a third ASM, although concerns regarding whether such a comparison would be ethical might persist. Perhaps a more feasible next step for future trials would be designing a prospective multicenter observational cohort study. Such data could be captured by a multicenter registry with standardized measures (including developmental outcomes and reporting for adverse effects). A registry spanning large geographical areas and reporting observational data would offer other important advantages: 1) given the relatively small number of infants undergoing surgical interventions, gathering data across multiple centers would offer important improvements to detect efficacy and harms; 2) such a registry could facilitate consensus about how to measure outcomes and 3) provide a framework for prospectively collecting data. Existing consortiums could play a role in facilitating development.

Development of core outcomes specific to infants could also support these efforts. Important outcomes identified by key stakeholders during protocol development included seizure freedom, seizure frequency, seizure severity, Engel classification, all-cause mortality, hospitalization, neurodevelopmental outcomes, quality of life, sleep quality, caregiver quality of life, treatment cost, and other adverse events. Given the range of seizure etiologies and surgical interventions, future studies should not only report these outcomes, but report outcomes separately for different seizure etiologies (i.e., HME vs. focal cortical dysplasia) and surgeries (i.e., focal cortical resection vs. frontal lobectomy). Without this level of detail, future systematic reviews are likely to encounter difficulty in drawing conclusions about specific etiologies or procedures.

Finally, as many seizure etiologies are relatively rare, families often face a challenge in identifying and then accessing a provider with clinical expertise. Furthermore, many infants with epilepsy also require other medical interventions, therapies, or services which may be challenging for families to obtain depending on their geographical proximity to specialized care or healthcare coverage. These and other factors may contribute to economic hardship after diagnosis. However, using telehealth and expanding coverage could not only improve the number of families with access to specialized expertise, but facilitate larger clinical trials to assess efficacy and long term outcomes.

This review does not provide cost information.

Applicability

As we did not stratify seizure type or etiologies, there may be differences between the patient population within this review and the overall infantile epilepsy population. Certain etiologies (i.e., particular genetic causes) may not have been captured by this review as their specific conditions may not lead them to enroll in trials. Although the scope of our systematic review did not include infantile spasms, three studies reported infants as having "epileptic spasms" which could include infantile spasms but also other seizure types of interest. Recent changes to nomenclature used to describe infantile spasms may have played a role. One study by Jackson et al. (2017) enrolled many patients with "epileptic spasms" which may have potentially been infantile spasms. We included these studies/infants, however, results may be less applicable depending on what proportion of study participants had infantile spasms.

We considered whether infants in included studies had similar baseline seizure frequency or seizure severity compared to infants with epilepsy in the general population. However, only 5 of 41 studies reported baseline seizure frequency, and only 3 of 41 studies reported baseline seizure severity. Thus, the evidence does not permit an assessment of applicability for either measure of disease severity.

A key issue with the applicability of the evidence on pharmacologic treatments concerns the use of concomitant medications. For pharmacotherapy, 11 of 15 studies enrolled infants who had already attempted other ASM, and they continued taking ASM even after the introduction of the medication under study. Studies reported many seizure types and syndromes, with few restrictions on enrollment, with one exception. The study by Grinspan et al. (2018) only enrolled patients with nonsyndromic epilepsy, and thus its result do not apply to infants with epilepsy syndromes. Thus, most of the evidence applies to the addition of another ASM to an existing regimen for the treatment of any form of epilepsy. We also note that the study by Piña-Garza et al. (2008) required a 40% response to lamotrigine in order for an infant to enter the randomized portion of the trial, and so its harms results only apply to lamotrigine responders.

Studies of dietary and surgical interventions are primarily relevant for infants with intractable epilepsy unresponsive to ASMs. As these interventions also require access to specialized expertise that is not widely available, not all infants and their families may have access. With regard to dietary studies, although studies often used the same Johns Hopkins protocol, differences in how the diet was implemented could affect applicability. For example, three studies did not use an initial fasting period, with one study switching from fasting to nonfasting in the middle of the study (perhaps because fasting is less feasible for this young population). In addition, studies aimed for different lipid to nonlipid ratios and had different diet schedules. Second, both KD and MAD require significant effort from parents, especially during the later stages when the patient transition from hospital to home. The considerable effort required to maintain compliance with dietary protocol, may make them less feasible for certain families. Finally, among 8 studies, only one was conducted in the United States and one in Europe. The remaining five were conducted in Asia and Egypt. Different diets and ASM availability could influence infant diets and affect applicability of these dietary interventions.

For surgical interventions, most studies were single center studies, each likely reflecting outcomes from a small number of surgeons. In addition to potential differences in surgical techniques across surgeons, studies reported a range of variations which, in some cases, reflected developments in the field. The most notable of these was a shift away from anatomical hemispherectomy to functional hemispherectomy or hemispherotomy over time. Our review included studies reporting procedures performed over four decades (from 1979 to 2020). Given changes in surgical technique and clinical care, findings may be less applicable to infants undergoing surgery today. Finally, we included a study assessing surgery for infants with epilepsy due to tumor; however, care for these infants and their outcomes will differ in important ways from other infants undergoing surgical resection.

Strengths and Limitations

One strength of this review is the exhaustive search for any evidence for interventions in infants 1 month to <36 months. Given early concerns about potential for sparse or insufficient evidence, we considered several strategies and altered inclusion criteria to allow alternative study designs and smaller studies. In addition, we removed a requirement that pre/post studies measure seizure frequency in the context of a prospective trial (i.e., not a chart review). This criterion was meant to address concerns about potential inconsistency and bias in seizure counts assessed outside the context of a formal study. However, had we used this criterion, nearly all surgical studies of effectiveness would have been excluded.

In addition, if abstracts only mentioned including children or pediatric populations, we screened the full text to ensure we captured any studies reporting data on age 1-36 months. We ultimately excluded over 1100 full articles that did not report any such data. However, these efforts identified 13 additional studies. Thus, almost a third of our evidence base (13 of 41 studies) could only have been identified with this level of scrutiny.

One limitation of this review is the exclusion of evidence published prior to 1999. Older studies may have assessed the efficacy in infants of some older ASM such as valproate, carbamazepine and phenytoin. Similarly, these criteria may have excluded older studies of procedures such as anatomic hemispherectomy or other procedures. However, including older studies would have raised potential challenges for applicability given changes in diagnosis and clinical care, and many of the newer ASM were not available earlier. Similarly, for surgical interventions, since some studies included patients extending back to the 1970’s, changes in surgical technique and clinical care may limit generalizability of findings. Importantly, we note that the lack of included evidence on older medications and procedures does not mean clinicians should necessarily exclude these treatments from consideration when tailoring treatments for individual patients. Our report is intended to provide a rigorous assessment of all available evidence published during this time frame.

Our inclusion criteria for this review may be considered a strength or a limitation, depending on judgements about relevance. Any evidence review must strike a balance between including important/relevant studies and excluding misleading/irrelevant studies, but no objective threshold exists. Key inclusion criteria for this review involved patient age (at least 80% must have been age 1-36 months at the time of treatment) and study size (n ≥ 10 for RCTs of any treatment, n ≥ 10 for non-randomized studies of surgery, and n ≥ 30 for non-randomized studies of medications or diets). Many have noted that this age group is clinically distinctive from both neonates and older children. Thus, some could argue our criteria were too lenient because we included studies that mixed this age group with others. Conversely, others might argue the criteria were too strict as some studies barely missed a numerical threshold (e.g., we excluded Arzimanoglou et al. (2019)205 as only 68% of patients were age 1-36 months). Notably, our pre-protocol criteria were stricter: ≥ 85% age 1-36 months, and n ≥ 30 for any study of medications or diet. During screening, we relaxed these criteria to include more studies we deemed sufficiently relevant, such as RCTs of medications or diets enrolling 11-29 infants.

Regarding etiology and seizure types, the scarcity and quality of evidence did not permit an examination of how these factors may influence treatment effectiveness. Seizure prognosis depends heavily on underlying seizure syndrome and etiology, along with comorbidities and concurrent therapies (such as number of ASMs). However, studies reported these data inconsistently. In addition, medical care and diagnosis has significantly evolved over the time periods captured in these studies. These factors precluded summarizing evidence on any particular etiology or seizure type as well as quantitative synthesis.

In addition to the evidence gaps noted earlier, no studies reported data on the cost of treatments. From stakeholder interviews, we learned that the cost of the intervention was an important factor for parents of young patients. While the review may be helpful to guide clinicians on their decision making, it provides no information on this important factor.

Evidence Gaps

We note several important evidence gaps. First, few studies have assessed treatments for children with epilepsies age <36 months. Despite lenient inclusion criteria (allowing less rigorous study designs, and including surgical studies with only 10 infants per treatment arm), we only identified a small evidence base. For example, although many different drugs and combinations of drugs are currently used in this population, we only identified evidence for five drugs of which only levetiracetam had evidence sufficient to conclude effectiveness. Furthermore, many studies (particularly for surgical interventions) were only included because a subgroup analysis reported on infants meeting criteria. Such studies often did not report clinical information (i.e., baseline seizure frequency, concomitant treatments, seizure etiologies) for the subgroup of interest.

Second, studies primarily reported on seizure frequency and seizure freedom, but largely failed describe other important outcomes such as hospitalization, neurodevelopment, infant quality of life, sleep outcomes, functional performance, and caregiver quality of life. In fact, only three studies described developmental outcomes (for surgical interventions) and only one study reported on functional assessment. This focus on seizure outcomes may be related to external pressures (i.e., focus on outcomes needed for drug approval) or perhaps ease of measurement. However, these outcomes are important to parents, caregivers and clinicians and reflect an important evidence gap.

Third, the most common study design in this literature was a single-arm study, and authors typically attributed outcomes (e.g., seizure freedom) to the study treatments alone, rather than other possible explanations (e.g., other treatments, spontaneous remission, short follow-up time). This attribution may be more reasonable for surgical studies (as compared to pharmacologic or dietary therapy studies) as these patients have typically failed pharmacologic and/or dietary treatments prior to undergoing surgery. However, in general, future clinical studies could benefit from greater awareness of alternative explanations of observed outcomes when considering trial designs.

Finally, we note the lack of evidence addressing several treatments including commonly used medications such as oxcarbazepine, as well as newer modalities including cannabidiol, neuromodulation and gene therapy. Our searches did identify pediatric studies of cannabidiol, vagus nerve stimulation, transcranial direct current stimulation, and responsive neurostimulation, but none met our inclusion criteria (typically these studies enrolled older children). Some studies of genetic testing or genome sequencing of children with epilepsy have been conducted, and future work may elucidate whether such testing improves outcomes through the optimal selection of treatment. Some gene therapy trials may be published soon; TSHA-105, which is an investigational gene therapy for a rare form of epilepsy called SLC13A5 deficiency, received EU orphan drug status in August of 2021, and clinicaltrials.gov lists a trial begun in March of 2021 (https://clinicaltrials.gov/ct2/show/NCT04798235 ).