Refractory seizures: Try additional antiepileptic drugs (after two have failed) or go directly to early surgery evaluation?


Address correspondence to Dr Patrick Kwan, Department of Medicine and Therapeutics, Prince of Wales Hospital, Shatin, Hong Kong. E-mail:


The goal of antiepileptic therapy is to achieve long-term seizure freedom with minimal or no adverse effects. Current evidence suggests that in many patients who have failed two appropriate antiepileptic drugs (AEDs) because of lack of efficacy, the chance of subsequent seizure freedom with further drug manipulation is low (reports ranging from as little as a few percent to nearly one-fourth of patients). Achieving this may require repeated drug manipulations. Surgery, in appropriately selected candidates, may render up to 70% of patients seizure-free when temporal resection is done, although frontal resection may have only a 25% yield. Both courses of actions (further drug trials and surgery) are associated with a plethora of potential adverse outcomes. Therefore, it is recommended that such patients be promptly referred to an epilepsy center for a comprehensive review of the diagnosis and management, which may include initial evaluation for surgery. Because presurgical evaluation and surgery itself may entail discomfort and risk, the decision to offer surgical treatment requires individual risk–benefit analysis that includes an assessment of possible success with additional trials of medication.

Despite appropriate drug treatment, one-third of patients with epilepsy continue to have seizures (Kwan & Brodie, 2000). For some, surgery may be a treatment option (Spencer & Huh, 2008). In particular, the effectiveness of anterior temporal lobectomy for drug-resistant temporal lobe epilepsy was confirmed in a randomized controlled trial (Wiebe et al., 2001) and is recommended by the American Academy of Neurology (AAN) in appropriately selected patients (Engel et al., 2003). However, it is generally thought that potential patients are often not referred for surgical evaluation, and even those who undergo surgery do so after considerable delay because of the difficulty in recognizing medical intractability early in the course of the epilepsy (Engel, 2008). Indeed, the number of drugs that need to fail before patients should be considered for surgery remains controversial (Kwan & Brodie, 2006). This is because the natural history of epilepsy, in particular its progress in response to treatment, remains poorly documented to enable reliable prediction of continued pharmacoresistance in individual patients (Kwan & Sander, 2004). The situation is compounded by the rapid development of antiepileptic drugs (AEDs) in recent years, with global approval of at least 13 new agents in the last two decades. Clearly, no patient will be able to try all AEDs in all combinations.

Given this dilemma, it has been common practice (Wiebe et al., 2001; Spencer et al., 2005; Bien et al., 2006) and recommendation (Kwan & Brodie, 2006) that patients should be considered for surgery after failure of two appropriate AEDs. Any such practice or recommendation must be based on a careful assessment of the probability of subsequent remission with further AED trials compared with surgery, balanced against the potential risks associated with either course of action. In this article, we review the evidence base for making an informed assessment and provide practical recommendation for individualized clinical decision making in patients who have been unsuccessfully treated with two AEDs.

Benefits of Continued Medication versus Surgery

Seizure freedom

The goal of antiepileptic treatment is long-term complete seizure control without adverse effects (Duncan et al., 2006). Long-term observational studies including newly diagnosed patients at the point of treatment initiation represent valuable sources of evidence to gauge the chance of seizure control with successive AED therapy, but few have specifically addressed the relationship between outcome and course of AED treatment. In one such study, among 470 patients who were previously never treated, 47% became seizure-free on their first drug, 13% on the second drug, but only 4% on the third drug or a combination of two drugs (Kwan & Brodie, 2000). Therefore, the probability of seizure freedom diminished progressively with successive AED regimens, whether substitution or add-on therapy, particularly after failure of two regimens (Fig. 1). Similar results were obtained in the analysis of the expanded cohort of 780 newly diagnosed patients from the same center: 50% became seizure-free with the first monotherapy, another 7% responded to the second monotherapy, and only 7% became seizure-free with the subsequent monotherapies or with polytherapy (Mohanraj & Brodie, 2005, 2006). In both analyses, patients who failed to respond to treatment due to lack of efficacy had worse outcome compared with those whose treatment failed for reasons such as adverse effects. These observations in mainly adult patients suggest that few become seizure-free after failure of two AEDs, particularly due to lack of efficacy.

Figure 1.

Probability of seizure freedom in patients with newly diagnosed epilepsy, according to the number of antiepileptic drug (AED) regimens. Dotted lines represent 95% confidence intervals (CIs) (From Brodie & Kwan, 2002).

Instead of following patients at diagnosis/treatment initiation, several recent studies recruited patients with chronic epilepsy who have already failed one or two AEDs. In an observational study of 155 adults who had previously failed one or more AEDs, 23% became seizure-free for 12 months or more after further drug trials, although it took up to six trials for some (Luciano & Shorvon, 2007). Observations from other adult cohorts who had failed at least two AEDs previously showed that subsequent seizure freedom (for at least 12 months) occurred upon further drug manipulation in approximately 4–5% of patients per year (Callaghan et al., 2007; Choi et al., 2008a), but with a probability of seizure relapse of 44% within 3 years (Choi et al., 2008a). These data suggest that in patients who have failed two or more AEDs, seizure freedom may still be attained in a small proportion, but it may involve repeated drug manipulations over considerable time, and may be temporary in nearly half of those who enter remission. Certain clinical features may help determine prognosis for remission with further medical treatment (Callaghan et al., 2007). Patients with an idiopathic etiology and epilepsy duration of less than 10 years had a greater chance of attaining remission. In contrast, patients with cryptogenic or symptomatic epilepsy and a longer duration of disease were more prone to experience continued seizures.

A fluctuating or remitting-relapsing course might be particularly common in childhood-onset epilepsy. In a prospective cohort of 140 children who had failed to respond to trials of at least two different AEDs considered appropriate for their seizures and type of epilepsy, some subsequently experienced repeated remissions and relapses, and only a small proportion became seizure-free for each of the additional drugs tried (Berg et al., 2009).

It is remarkable that within the cohort of 246 patients observed by Callaghan et al. (2007), 21 received surgery and 11 (52%) became seizure-free, compared with a seizure-free rate of only 11% in the rest of the cohort who were treated medically. The superiority of surgery over continued medication in terms of seizure control at short term was demonstrated conclusively by Wiebe et al. (2001). In this landmark study, 80 patients with temporal lobe epilepsy who failed treatment with two or more AEDs were randomized to surgical (in the form of anterior temporal lobectomy) or medical treatment. At one year, 58% of patients in the surgical group were free of complex partial or secondarily generalized seizures compared with only 8% in the medical group.

Because of ethical reasons, it would be difficult to conduct such randomized trials with prolonged follow-up; however, long-term data are available from observational, albeit mostly uncontrolled, studies. In the prospective Multicenter Study of Epilepsy Surgery, of 339 patients followed for 2–7.3 years after resective surgery (mainly temporal lobectomy), 66% were seizure-free for at least 2 years (Spencer et al., 2005). Outcome showed a trend to be better in patients who had mesial temporal lobe resection (68% seizure-free) than those who had extratemporal neocortical resections (50% seizure-free). A nonrandomized, controlled study analyzed outcomes of 242 patients with temporal lobe epilepsy who had failed at least two AEDs seen at a single epilepsy center (Bien et al., 2001). After a mean follow-up of approximately 5 years, 44.6% of those who eventually underwent surgery (n = 148) were seizure-free during the 12 months before follow-up compared with only 4.3% of those treated by AEDs alone (n = 94).

Long-term studies also suggest that temporal lobe resections were often associated with a good outcome. Téllez-Zenteno et al. (2005) performed a meta-analysis of 83 published studies including 7,343 patients undergoing resective or nonresective epilepsy surgery and followed for a mean/median of at least 5 years (Table 1). The median proportion of seizure freedom at the last reported follow-up was higher (66%) with temporal lobe resections. In contrast, although many patients benefit, results were less impressive after occipital or parietal (both 46% seizure-free) or frontal (27% seizure-free) lobe resections, as well as nonresective operations (16% seizure-free after multiple subpial transections and 35% free of most disabling seizures after callosotomy). Among patients undergoing temporal lobe resections, those with discrete abnormalities identified preoperatively, such as hippocampal sclerosis and other foreign tissue lesions, had a higher probability of seizure freedom than those without obvious abnormality (Mcintosh et al., 2004). Indeed it has been consistently shown that patients with hippocampal sclerosis as the pathologic substrate have the best prognosis after temporal lobe resections (Radhakrishnan et al., 1998; Mcintosh et al., 2004; Tonini et al., 2004; Spencer et al., 2005), but within this group, no reliable predictive clinical factor (Kilpatrick et al., 1999; Hennessy et al., 2001; Hardy et al., 2003; Janszky et al., 2005; Aull-Watschinger et al., 2008) or biologic marker for long-term outcome has been identified. The preceding data indicate that temporal lobe surgery has a reasonably high probability of providing seizure relief, particularly when the preoperative magnetic resonance imaging (MRI) discloses a lesion. Extratemporal surgery, however, results in remission less often. Frontal lobe surgery, in particular, produces an excellent outcome far less often than desired. Hence, it would appear that temporal lobe surgery is superior to continued medical therapy after failure of two drugs. However, the case for superiority of frontal lobe surgery to continued drug therapy after failure of only two medications is not overwhelming. In this case, the relative risks of each course of treatment might serve as a guide for the most appropriate choice. This topic is addressed later in this review.

Table 1.   Meta-analysis of long-term seizure freedom after different types of epilepsy surgery
Type of surgeryNo. of patients (studies)% seizure-free at 5 years (95% CI)
  1. Adapted from Téllez-Zenteno et al. (2005).

  2. CI, confidence interval.

Temporal3895 (40)66 (62–71)
“Extratemporal”169 (2)34 (28–40)
Frontal486 (7)27 (23–30)
Parietal82 (1)46 (56–62)
Occipital35 (1)46 (29–63)

Similar as in medical therapy, relapse may occur in patients who had surgery after a considerable period of seizure freedom that may not be temporally related to AED withdrawal (Spencer et al., 2005). Sperling et al. (2008) reported that in a cohort of 159 patients who had been seizure free for at least 5 years postsurgery (88% had anterior temporal lobectomy), there was still a probability of seizure relapse of 4% per year (Fig. 2). However, at the last follow-up of the study (mean follow-up duration 12.2 years), 89.9% of patients had been seizure free for at least the previous year, suggesting that seizure relapses in these patients are often isolated events and that seizure control can be regained in the majority. Nonetheless, the notion that surgery produces a “cure” in a majority of patients is not substantiated by the data. Although surgery leads to prolonged periods of remission, most patients experience relapses after surgery.

Figure 2.

Kaplan-Meier graph showing annual rate of occurrence of first seizure after resective epilepsy surgery in patients who had been seizure-free for the first 5 years. (From Sperling et al., 2008).

Finally, one other advantage might be offered by surgery. It offers the possibility for some patients to reduce or stop AED therapy, which may be associated with a sense of “cure” by the patients. However, only a minority of seizure-free patients stop or reduce their medication successfully (Berg et al., 2006). Moreover, the regular occurrence of postoperative relapses might lead many physicians to err on the side of caution and advise continuation of therapy.

Mortality and quality of life

The pursuit of ineffective therapy is encumbered not only by the risks of that therapy, but also by the risk of continued seizures. Significantly, seizure freedom after surgery is associated with decreased mortality and improved quality of life. In a cohort of 583 patients undergoing epilepsy surgery, those with recurrent seizures after surgery (n = 325) had a higher standardized mortality ratio [SMR = 5.75, 95% confidence interval (CI) 3.51–9.27] compared to those who were seizure-free (n = 258, SMR = 0.45, 95% CI 0.02–2.94), implying that the excess mortality associated with refractory epilepsy is eliminated after epilepsy surgery when seizures are abolished (Sperling et al., 2005). A clinical trial to establish the putative positive effect of surgery on mortality rate would require long-term follow-up of a large number of patients and likely be prohibitively costly. Choi et al. (2008b) recently used a decision analytic model and predicted that for a 35-year-old patient with temporal lobe epilepsy inadequately controlled by at least two AEDs, surgical therapy would be associated with an increase in life expectancy of 5.0 years and 7.5 in quality-adjusted life-years. In addition, prospective studies have shown that absolute seizure freedom postsurgery is associated with a durable improvement in quality of life measures (Markand et al., 2000; Spencer et al., 2007).

Risks of Continued Medication versus Surgery

Any decision to recommend a particular therapy must account not only for its benefits, but also its risks. Both medical and surgical therapy pose specific risks, independent of the risk associated with seizures. This section will focus on the direct risks of these treatment approaches.

Risks of surgery

Neurologic surgery poses medical, neurologic, and psychiatric risks, in addition to the discomforts and social restrictions entailed. The surgical evaluation process and surgery itself typically cause many weeks of disability, with 4–8 weeks lost either from work or school, in addition to the disruption of ordinary home life. The most common medical complications include deep vein thrombosis, wound infections, as well as transient endocrine abnormalities, often noted by menstruating women. Psychiatric sequelae of epilepsy surgery are not uncommon, with a one-fourth or more of patients experiencing transient dysphoria, depression, or rarely, mania (Kohler et al., 2001). Although permanent serious neurologic complications are uncommon, minor and moderate sequelae are common, related to the proximity of the surgical site to vital cortex. In particular, dominant hemisphere resections put language at risk, and a resection near the central sulcus can jeopardize motor or sensory function. These risks are summarized in Tables 2 and 3.

Table 2.   Complications of temporal lobectomy
 Naming and fluency deficits following dominant hemisphere resection in naming; incidence ∼50%
 Depends on preoperative level of function, improves with time after surgery
 Verbal memory impairment after dominant temporal resection, depends on preoperative level of function
Cranial nerve palsy
 Affects primarily CN 4; incidence 1.5%
 Hemiparesis in 1–2% of patients
 Quadrantanopsia occurs in 7–50% of patients (technique dependent)
Table 3.   Complications of extratemporal resections
Frontal resection
 Broca’s aphasia
 Motor deficits: hemiparesis, clumsiness
 Supplementary motor syndrome: mutism, hemiparesis, and neglect (typically transient)
 Personality change
Parietal resection
 Aphasia (dominant hemisphere)
 Hemisensory deficit
 Visual field deficit (inferior quadrant field cut)
Occipital resection
 Visual field deficit
 Alexia (dominant hemisphere)

The chance of complications can generally be predicted in advance of surgery, and operations can be classified a priori as low risk or higher risk. For example, the patient is exposed to minimal risk if a nondominant frontal pole lesion must be excised. In contrast, the patient who requires a resection in cortex near Broca’s area or the motor strip has a significant chance of incurring a permanent deficit. Consequently, the decision to offer an operation must in part be based on the likelihood of producing neurologic sequelae, which should be predicted, and, therefore, avoided, as far as possible based on detailed preoperative investigations such as neuropsychological testing and noninvasive or invasive functional mapping. This affects not only postoperative function but also the chances that surgery will be successful, since surgery in a critical functional area is most apt to lead to incomplete resection of the epileptogenic focus.

Lastly, surgery poses one other potential risk. The neurosurgical procedure creates a new lesion in the brain, which at times may itself cause seizures. This has been demonstrated for other types of surgery. For example, routine shunt surgery in children is often associated with the later development of seizures (Bourgeois et al., 1999). Some patients experience seizure recurrence long after surgery has been performed, and it is possible that in the course of excising one epileptogenic lesion, another was produced.

Risks of medication

AEDs are not entirely benign. Adverse effects include both neurologic and nonneurologic symptoms, which are far too numerous to comprehensively list in this brief review (Asadi-Pooya & Sperling, 2009). Dose-related central nervous system (CNS) adverse effects are common to all medications; all have the potential to alter cognitive function, and cause sedation and imbalance, and many may influence mood or personality. Other side effects are specific to particular agents, such as weight loss or weight gain, tremor, and irritability, among others. Potentially life-threatening idiosyncratic reactions have also been reported with most drugs, including Stevens-Johnson syndrome, hepatic failure, and aplastic anemia. Antiepileptic medications are also teratogenic, posing real risk to the developing fetus; in an ideal world, women would not take any medication during pregnancy.

Long-term administration of AEDs, especially the enzyme-inducing agents, appears to carry added risk. Cytochrome P450 enzyme inducers have been associated with reduction in sex hormone levels, lower vitamin D levels, and elevated C-reactive protein (CRP) and cholesterol levels (Herzog et al., 2006; Pack et al., 2008; Mintzer et al., 2009). These effects can lead to increased risk of osteopenia and osteoporosis, menstrual irregularities, reduced fertility, sexual dysfunction, and increased risk of cardiovascular illness. Other long-term adverse effects may exist, which presently remain unknown. In addition, AEDs that induce the hepatic metabolic enzymes are prone to undesirable drug–drug interactions with a range of non-AEDs, including commonly used drugs such as oral contraceptive pills and warfarin, potentially leading to treatment failure of these agents (Perucca, 2006).

Hence, although most patients tolerate their medical therapy with no or minimal difficulty, the preceding concerns must be considered. Prolonged exposure to AEDs might be detrimental for some patients. Limiting the period of exposure is desirable, So implementation of a therapy such as surgery might offer advantage if medication could later be discontinued.

Recommendation and Conclusion

The overall goal of antiepileptic therapy is to achieve long-term seizure freedom as soon as possible with minimal or no adverse effects. Summarizing current evidence, in patients who have failed two appropriate AEDs because of lack of efficacy, the chance of subsequent seizure freedom with further drug manipulation is probably around 5–10% per year, with perhaps a longer-term remission rate of at most 25%. Moreover, any remission might not be durable but certain simple clinical features can help ascertain which patients are least likely to remit. In comparison, in appropriately selected candidates, up to 70% may become seizure-free long-term “immediately” after resective surgery. In this respect, given that most patients respond to the first two AED schedules (Kwan & Brodie, 2000; Mohanraj & Brodie, 2005, 2006), the key step in the management process for those who fail to do so is a comprehensive review of the diagnosis and management, preferably at an epilepsy center where initial evaluation for surgery may be made if appropriate. However, because presurgical evaluation and surgery itself may entail risks, the decision to offer surgical treatment requires individual risk–benefit analysis that includes an assessment of possible success with additional trials of AEDs. For instance, in situations where surgery can produce a high remission rate, as is seen after anterior temporal lobectomy or amygdalohippocampectomy for a lesion such as mesial temporal sclerosis or a benign neoplasm, it is advisable to recommend a surgical procedure. In contrast, the patient who has a normal MRI scan and a probable extrahippocampal seizure origin is not best served by immediate surgery after failure of only two medications. This patient will likely have a higher surgical risk, and surgery may have perhaps as low as a 25% chance of inducing long-lasting seizure remission. In this situation, further trials of AED therapy appear warranted before embarking on more extensive and possibly invasive surgical evaluation procedures.


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.

Disclosure: PK has consulted for and/or received research grants from Esai, Johnson & Johnson, Pfizer, and UCB Pharma. MS has consulted for Dainippon Pharmaceuticals and Valeant, and is on the speaker’s bureau for UCB and Pfizer.