Evidence reviews for surgery: referral and surgical interventions

1.2.1. Summary of the protocol

Table 1

PICO characteristics of review question.

1.2.2. Methods and process

This evidence review was developed using the methods and process described in Developing NICE guidelines: the manual. Methods specific to this review question are described in the review protocol in appendix A and the methods document. Declarations of interest were recorded according to NICE’s conflicts of interest policy.

1.2.3. Effectiveness evidence

No relevant studies were found

1.2.4. Economic evidence

1.2.5. Economic model

This area was not prioritised for a new cost-effectiveness analysis.

1.2.6. Unit costs

Relevant unit costs are provided below to aid consideration of cost effectiveness.

Table 2

Costs of pre-surgical evaluation tests.

Costs for epilepsy surgery were found by looking up OPCS codes for the epilepsy surgery types listed on the review protocol. These were then linked to the HRG codes using the HRG4 reference costs grouper ‘code to group’ spreadsheet. A single OPCS code can be linked to several HRG codes depending on whether certain ‘flags’ are raised that changes the complexity of the procedure. All the codes HRG codes identified are listed below for an illustration of the costs.

Table 3

Costs of surgery.

Table 4

Anti-seizure medication costs.

1.2.7. The committee’s discussion and interpretation of the evidence

No evidence was found examining the use of different criteria for referral or surgery. The committee, therefore, agreed to use the clinical and health economic evidence on the effectiveness of surgical procedures to inform their recommendations regarding referral. The evidence and discussion are in section 1.3.

1.3.1. Summary of the protocol

For full details see the review protocol in Appendix A.

Table 5

PICO characteristics of review question.

1.3.2. Methods and process

This evidence review was developed using the methods and process described in Developing NICE guidelines: the manual. Methods specific to this review question are described in the review protocol in appendix A and the methods document.

For the outcome of seizure freedom, hazard ratios (HRs) for the first seizure were either not available from the papers or poorly reported. HRs were therefore calculated from Kaplan Meier survival graphs and other data provided in the studies: life tables were constructed by the reviewer, and HRs for the first seizure were calculated using excel.

Declarations of interest were recorded according to NICE’s conflicts of interest policy.

1.3.3. Effectiveness evidence

1.3.4. Summary of studies included in the effectiveness evidence

Table 6

Summary of studies included in the evidence review.

See Appendix H for full evidence tables.

1.3.5. Summary of the effectiveness evidence

Table 7

Clinical evidence summary: Surgery versus medical/ waitlist-control.

See Appendix J for full GRADE tables.

1.3.6. Economic evidence

1.3.7. Excluded studies

Four economic studies^{9, 44, 50, 82, 12}relating to this review question were identified but were excluded due to methodological limitations and the availability of more applicable evidence. These are listed in Appendix G, with reasons for exclusion given.

See also the health economic study selection flow chart in Appendix D.

1.3.8. Summary of included economic evidence

Table 8

Health economic evidence profile: Testing strategies following discordant EEG and MRI findings versus medical management.

Table 9

Health economic evidence profile: Intracranial EEG (subdural grid electrodes) versus iEEG (stereoelectroencephalography) versus medical management.

1.3.9. Economic model

An original cost-utility analysis was developed, assessing the cost-effectiveness of resective epilepsy surgery in adults with drug refectory epilepsy. Original health economic modelling was also planned to model the cost-effectiveness of resective epilepsy surgery in children, but insufficient data were available to model for this population. Full details of the health economic analysis can be found in the Economic analysis report.

The committee identified this as a high priority area as they thought that currently, there could be a reluctance to refer people for resective epilepsy surgery. The committee wanted to evaluate the benefits of resective epilepsy in terms of improved seizure freedom and long-term cost savings.

Model structure

The following comparators were included in the analysis:

1. Resective epilepsy surgery
2. Medical management

The population of the analysis was adults with drug refractory epilepsy.

A two-part model was developed, which included a decision tree to model post-procedural outcomes (over 1 year) followed by a Markov model for the estimation of quality-adjusted life-years and costs over the lifetime of the patient. The decision tree structure can be found in Figure 1, and long-term Markov model structure can be found in Figure 2.

Figure 1

Decision tree.

Figure 2

Markov model.

The model base case analysis was built probabilistically to take account of the uncertainty around input parameter point estimates. A probability distribution was defined for each model input parameter. When the model was run, a value for each input was randomly selected simultaneously from its respective probability distribution; mean costs and mean QALYs were calculated using these values. The model was run repeatedly – 5,000 times - and results were summarised.

Assessment for resective surgery survey

A comprehensive survey was administered to participating adult epilepsy surgery centres to obtain the average number of tests for people undergoing assessment for resective epilepsy surgery. Ten surgical centres submitted data for a total of 762 people.

Overall, fourteen epilepsy surgical centres were contacted, resulting in a response rate of 71%. The committee was provided with a list of the participating surgical centres and concluded the data would provide a representative sample to obtain the resource use for preoperative assessment.

The mean number of tests are reported in Table 10.

Table 10

Preoperative assessment cost.

The centres were also asked about the outcome of patients being assessed for surgery to assess what proportion of those being assessed for surgery proceeds to have a surgical resection. The probability of being a surgery candidate was determined to be 41.3% across all centres. In the model, we add the test of costing those patients that did not go on to have surgery as well as the cost of the surgical patient themselves. So, the total assessment cost per patient undergoing surgery was £8,182 + £11,628 = £19,809, where £11,628=£81,812*(58.7%)/41.3%.

Cost-effectiveness Results

The base case probabilistic model results indicated surgery was cost-effective at NICE’s £20,000 threshold with a cost per QALY of £11,425. The total cost for surgery was higher compared to medical management (£56,204 compared to £31,627), but the total QALYs were also higher for surgery (15.91 compared to 13.76). The higher cost for surgery was largely driven by the high cost of assessment for resective epilepsy surgery and procedure costs. Greater QALYs for surgery were obtained because more people receiving resective epilepsy surgery achieve seizure freedom compared to those receiving medical management, and higher standardised mortality ratios are associated with people who are not seizure-free.

Table 11

Base case cost effectiveness results (probabilistic).

The sensitivity analyses showed that the results were a little sensitive to the utility values, and costs, but only when the time horizon was lowered (to 15 years) did the cost per QALY gained exceed the £20,000 per QALY gained threshold - Table 12. Only when all the most pessimistic assumptions were made did the cost per QALY gained exceed £30,000 per QALY gained.

Table 12

Sensitivity analysis (deterministic).

1.3.10. Unit costs

Please see unit cost presented in section 1.2.6.

1.3.11. Evidence statements

1.4.1. The outcomes that matter most

The most important outcome was agreed by the committee to be quality of life, as this encapsulates the effects of treatment most relevant to the person with epilepsy. Other outcomes of similar importance were mortality, seizure recurrence, serious adverse events and cognition. Mortality and seizure recurrence were regarded as highly relevant outcomes to evaluate the potential harms of not using surgery, whilst serious adverse events and cognition were deemed the outcomes best suited to measure the harms of surgery itself. Although serious adverse events and cognition had originally been deemed non-critical outcomes at the protocol stage, it became apparent during discussion of the evidence that these were centrally important to decisions concerning recommendation of surgical intervention, because they could have a significant impact on function. All other outcomes were deemed important but less liable to affect recommendation decisions.

1.4.2. The quality of the evidence

The evidence included for this review was for resective surgery only. Quality of the evidence ranged from ‘high’ to ‘very low’. The most common reasons for downgrading of ratings were imprecision and risk of bias. The committee noted that imprecision was related to a lack of power in the included studies, which had relatively small sample sizes. The risk of bias was usually related to a lack of blinding (2 studies) or lack of allocation concealment (1 study). Given the nature of the interventions, surgery versus waiting list control, lack of blinding was inevitable. Therefore, the committee agreed that despite the risk of bias ratings, the studies were well conducted, and the overall quality was good. The committee was relatively confident of the validity of the evidence and supported a strong recommendation for surgery in both adults and children.

1.4.3. Benefits and harms

The committee agreed that surgery led to much better improvements in quality of life than medical care and that this was a crucial benefit of surgery for both children and adults with drug=resistant epilepsy. This benefit was generally of large magnitude, suggesting an effect that would be noticeable and important to the person with epilepsy, and it was also highly unlikely to be due to chance (sampling error). The committee also agreed that the greater reductions in seizure recurrence resulting from surgery would be an important benefit to the patient, particularly in view of the large effect size observed from the time to event data. Although surgery’s relative effects on mortality were small and consistent with possible sampling error, the committee stated that the single death occurring in the medical care group was made more significant by virtue of it being: sudden and unexpected. This suggested it might be due to SUDEP, which was viewed by the committee as potential harm resulting from not providing surgery in a timely fashion.

The committee also weighed up the accompanying harms of surgery in terms of clear examples of greater cognitive deficits post-surgery in the studies. The committee questioned whether pre-surgical methods such as the sodium amobarbital procedure (Wada test) had been carried out in the studies to try to reduce the likelihood of these harmful effects occurring. The fact that these methods had not been used in the included studies suggested that the reported cognitive defects might not always be an inevitable result of surgery, as they might, in practice, be ameliorated to some extent by suitable pre-surgical assessment strategies.

The cognitive tests showing harms for surgery were discussed further in some detail by the committee. In relation to the ‘Boston Naming Test’, members of the committee explained how patients often find that deficits in this test do not always translate to dysfunction in everyday life, and that the majority of patients would accept living with a reduction in naming capacity. Furthermore, it was discussed how there are approaches that may be used to minimise any disability caused by naming difficulties. These might include standard interventions offered by speech and language therapists and neuropsychologists, such as phonologic and orthographic cues, semantic feature analysis, contingency-based cueing hierarchies, and repeated conversational engagement. With respect to the ‘delayed recall test’, the committee accepted that the reported deficits were a more intractable problem, as there were few ways of avoiding them. It was described how such deficits are manifested in about a third of cases by a cognitive ‘aging’ of around 10 years. However, the committee stressed that the risks of cognitive decline when having surgery were comparable to the risk of SUDEP when not having surgery. Given that cognitive decline is less serious than death, the committee agreed that the benefits of surgery were not negated by the evidence of cognitive decline.

The other serious adverse events observed in the studies, such as motor deficits, were also discussed, but the consensus was that these largely self-limiting effects did not shift the overall balance of benefits and harms away from an overall benefit for surgery. The committee acknowledged the importance of counselling as part of a surgical workup to discuss the balance between risks and benefits of surgery with patients or their carers to enable informed decision making.

The committee also discussed how the balance of benefits and risks depends on the complexity of the surgery, with more complex surgery leading to less benefits and more risks. In relation to this, the committee discussed how paediatric epilepsy surgery was very often more complex than adult surgery. For example, paediatric surgery would often involve more extratemporal surgery, as well as potentially risky procedures such as hemispherotomy or corpus callosotomy. However, it was agreed that even in very complex paediatric surgery the balance of benefits and harm from surgery would often still be superior to that of medical care.

The committee noted that no evidence had been found for surgery in people with learning disabilities and discussed that this population are less likely to have surgery. This may happen because of difficulties in gaining consent, or they are not referred due to a belief they would be unable to cope with the surgical assessment. In addition, the committee referred to evidence that this group might have poorer outcomes from surgery because of intractable brain pathology. The committee also noted people with genetic abnormalities may also sometimes be excluded from referral for a surgical assessment, and agreed that people with learning disabilities and those with genetic abnormalities should be considered for surgical assessment when it is indicated.

The committee agreed that any referral for surgery should be made as soon as a person had been identified as appropriate for surgery. It was agreed that early referral would avoid unnecessary risks resulting from further seizures during a waiting period, without any attendant benefits from such a delay. The committee discussed the reasons why delays tend to occur before referrals are made, even when a patient is clearly suitable for surgery. General misunderstanding of what surgery can offer was offered as a major reason, and it was discussed how improved education of clinicians was important.

The committee was therefore unanimous that the overall clinical benefits of surgery should make it an option for everyone – both adults and children - with drug-resistant epilepsy i.e., has tried two or more ASMs and is experiencing at least 2 seizures a month. The need to make decisions based on the individual patient’s characteristics and wishes was stressed, but it was also discussed how all patients with drug-resistant epilepsy should be given an opportunity to be referred to a tertiary centre that could consider a surgical strategy for the patient. The committee agreed that an adult with drug-resistant epilepsy should be referred to a specialist tertiary centre, and a child should be referred to a tertiary paediatric neurology service, both for consideration of surgical treatment, as early as possible.

1.4.4. Cost effectiveness and resource use

Two economic evaluations were identified for the review question assessing the cost-effectiveness of resective epilepsy surgery (Burch 2012 & Kovacs 2021). No economic evaluations were identified for the review question assessing the criteria for assessment for resective epilepsy surgery.

Burch 2012 assessed the cost effectiveness of testing strategies for assessment for resective epilepsy surgery following discordant EEG and MRI findings from a UK NHS and personal social services perspective. They compared three strategies:

1. Medical management (MM)
2. Patients received FDG-PET.

   1. if the results of the preoperative assessment test were positive (e.g., the epileptic zone was located, and resective epilepsy surgery was feasible), patients received surgery.
   2. If the results of the FDG-PET were negative or uncertain, patients received MM.
3. Patients first received FDG-PET.

   1. Then, as in intervention 2, if the results were positive, patients received surgery,
   2. if the results were negative, patients received MM.
   3. where the results of the test were uncertain patients received an iEEG.

      1. A positive iEEG resulted in surgery being undertaken and
      2. a negative or uncertain result led to continued treatment with MM.

For FDG-PET compared to MM the cost was £1,671 per QALY gained, and for FDG-PET +iEEG compared to MM, it was £1,925 per QALY gained.

Kovacs 2021 also assessed the cost-effectiveness of three different strategies from a Hungarian health care perspective. The cost-utility analysis undertaken by Kovacs 2021 was based on the cost-effectiveness analysis undertaken by Burch 2012, with differences in some of the data inputs used to populate the model and the preoperative assessment tests being evaluated. In Kovacs 2021. MM was compared to two different types of iEEG monitoring – placement of subdural grid electrodes (SDG) and stereotactic implantation of depth electrodes (SEEG). If localisation of the epileptic zone was identified, resective epilepsy surgery would be conducted. However, if the iEEG was unsuccessful in identifying the epileptic zone patients would receive MM. SDG cost an extra £2,802 compared to MM and SEEG cost £4,284 per QALY gained compared to MM.

Neither of these economic evaluations captured the RCT evidence identified in the clinical review. In addition, the committee noted that some of the data inputs used in these studies might not reflect the majority of epilepsy surgery patients. The target population in these studies would typically be people where the epileptic zone is more difficult to localise as the preoperative assessment tests being evaluated would normally be conducted at the latter stages of the assessment for resective epilepsy surgery pathway. The committee acknowledged that when the epileptic zone is more difficult to localise – and subsequently the area of the brain to resect may not be as well defined – poorer outcomes post-surgery may be observed compared to people where the epileptic zone was identified through fewer preoperative assessment tests.

The health economic studies included in the review were limited to the cost-effectiveness of specific preoperative assessment tests; therefore, the costs of other preoperative assessment tests were not included in the overall assessment of the cost effectiveness.

Due to the high clinical and economic importance concerning the cost effectiveness of resective epilepsy surgery original health economic modelling was also undertaken to assess the cost effectiveness of resective epilepsy surgery in adults. Unfortunately, there was insufficient data to model the cost-effectiveness of resective epilepsy surgery in children. The lifetime cost of surgery was higher than for MM (£56,204 and £31,627 respectively) but the QALYs gained were also greater in the surgery arm (15.91 QALYs compared to 13.76 QALYs). Compared with medical management, surgery cost an extra £11,425 per QALY gained, which is below NICE’s £20,000 threshold.

The clinical evidence included in the review was included in our original health economic analysis; however, it only provided data to populate the data in our 1-year decision tree. This is because it is challenging to conduct long-term RCTs assessing the effectiveness of epilepsy surgery due to cross-over. Observational data was therefore required to inform the long-term effectiveness of surgery and MM. A large long-term observational study based on a UK population (de Tisi 2011) was used to populate the long-term effectiveness outcomes for people after epilepsy surgery. However, there were fewer data available for a drug refractory population who continue MM. These data were subsequently taken from Callaghan 2011, which evaluated the remission and relapse rate in a drug-resistant epilepsy cohort of 246 patients from the USA. In de Tisi 2011, data were reported up to 15 years and in Callaghan 2011 data were reported up to 5 years. To account for the potential uncertainties in the long-term (lifetime) data for both surgery and medical management, a sensitivity analysis was conducted using a 15-year time horizon. Although, at 15 years, there is still uncertainty over medical management because data in Callaghan 2011, was limited to 5-year follow-up. In this analysis, the cost per QALY gained was £28,231, which is above NICE’s £20,000 threshold but below NICE’s £30,000 threshold. However, the committee thought that this was conservative, and it is reasonable to assume that the impact of surgery can continue for longer than 15 years for most individuals.

In most long-term outcome studies (including de Tisi 2011) assessing epilepsy surgery, seizure freedom was defined as being completely seizure-free or with only simple partial seizures, now termed focal-aware seizures (FAS). This is reasonable but this definition did not correspond with the definition used in the trials or the studies that were sourced for health state utility scores and standardised mortality ratios, which were only people who were completely seizure-free. To overcome the challenges posed by these differential definitions, adjustments were made to the standardised mortality ratios (SMRs) and utilities for seizure freedom in the surgery arm using the proportion of people that experienced FAS in de Tisi 2011. The utility and mortality for people experiencing only FAS is not known, and so conservative assumptions were made, which if anything, might have under-estimated the benefits of surgery, but as only 18% of people of the seizure-free sample had experienced FAS, the committee concluded this would not alter the overall results of the cost-effectiveness analysis.

Callaghan 2011’s definition of drug refractory epilepsy was stricter than the current definition of drug refractory. Callaghan defined drug-resistant epilepsy as people who had failed on at least two antiseizure medications (ASMs) and were experiencing at least one seizure per month. The current ILAE definition of drug refractory epilepsy is the occurrence of uncontrolled seizures despite two tolerated and appropriately chosen ASMs. Therefore, the cohort of people in Callaghan 2011 may have had more severe drug refectory epilepsy compared to a drug-resistant cohort as defined by the ILAE definition. The committee did, however, note that the estimated proportion of people entering seizure freedom (5.6%) and relapsing (22%) each year seemed reasonable.

The committee discussed how the results of the adult epilepsy model may translate into a paediatric population. The committee discussed that the cost of pre-surgical evaluation may be more expensive for children as they might require additional tests. However, the committee noted seizure freedom after resective epilepsy surgery might be more likely in children than adults, and the benefits for children could be accrued over a longer period.^{33, 34}

There is also some evidence that if seizure-free they are more likely to be able to stop taking anti-seizure medication,^{33, 34} which would be cost-saving in the longer term. In addition, the committee noted children with drug refractory epilepsy are likely to have more outpatient appointments than adults. Therefore, rendering children seizure could result in additional downstream cost savings because less outpatient appointments are required for people rendered seizure-free. Additional studies have shown children with drug refractory epilepsy who receive surgery have, better cognitive development, better outcomes in school, and greater chances of employment in adult life compared to those who continue to receive MM.^{8, 64–66}

A lifetime re-operation rate of 4% was incorporated in the adult, model and the committee concluded this rate of 4% would likely be similar in a paediatric population. The committee did, however, note that for children undergoing resective epilepsy surgery, the need for re-operation may also arise in adulthood which would incur additional costs however, this is only for a small proportion of people. The committee concluded that the cost savings observed in a paediatric population would likely outweigh the additional costs that may be incurred from preoperative assessment and additional re-operation in adulthood and therefore concluded resective epilepsy surgery in children is highly likely to be cost-effective.

Overall, based on the results of the included health economic studies and the original health economic analysis, the committee concluded that resective epilepsy surgery in a drug refractory population is highly likely to be cost-effective in both adults and children. The committee discussed that although there are limitations with the evidence used in the health economic model, these were dealt with in the most appropriate way.

In current practice, epilepsies services for surgery are managed separately for adults and children. Epilepsy services for children are run by the Children’s Epilepsy Surgery Service (CESS), whereas epilepsy surgery for adults is managed at tertiary epilepsy centres. The CESS centres were developed to increase the levels of paediatric resective epilepsy surgery in England. Since CESS was developed, the number of children undergoing preoperative assessment and resective epilepsy surgery in children has increased. The committee noted that the target number of referrals for preoperative assessment set by CESS, are currently below target. Target levels for preoperative assessment and resective epilepsy surgery are predicted by CESS epidemiologically. The committee concluded the recommendations made would help CESS achieve their targets and therefore are not expected to result in a substantial resource impact.

Resective epilepsy surgery is provided for adults at tertiary epilepsy centres. However, the committee noted that levels of referral for epilepsy surgery are sub-optimal. The committee prioritised this area for original health economic modelling in the hope they could recommend everyone with drug refractory epilepsy be referred for pre-surgical evaluation.

The health economic model incorporated the cost of pre-surgical evaluation in the total costs of surgery (including the costs for people who were referred for surgery and underwent pre-surgical evaluation but did not go on to have surgery). Sensitivity analyses were also conducted, assuming a higher and lower cost for assessment for resective epilepsy surgery. This was calculated at the highest cost out of the nine participating centres and the lowest cost. When the higher cost was used, resective epilepsy was still cost-effective (£16,827 per QALY gained). The committee noted that this cost for assessment of resective epilepsy (£13,178) is likely more reflective of being undergoing more complex preoperative assessments.

Overall, as surgery was assessed to be cost effective the committee concluded they were able to make a strong recommendation to refer adults, children and young people with drug resistant epilepsy for assessment for resective surgery.

The committee noted that in current practice referral to an epilepsy surgery centre for people who are drug refractory can take years. This is due to several factors, such as the misconception of clinical uncertainty surrounding the efficacy of surgery or healthcare professionals taking a view that referral should be a ‘last resort’ once a large number of ASMs have been tried. The committee discussed that once a person has failed two appropriately chosen ASMs the chances of obtaining seizure freedom through use of ASMs diminishes significantly. The committee noted that referral to an epilepsy surgery centre to enter the assessment for resective epilepsy surgery pathway does not necessarily mean surgery will take place for the patient in question. A person may not be an eligible candidate for epilepsy surgery, or they may not wish to proceed with surgery. Given all of the data presented, the committee was, though, clear that people with drug-resistant epilepsy should be referred promptly to a tertiary epilepsy centre for consideration of epilepsy surgery.

This recommendation may lead to a substantial resource impact as more adults will likely be referred for assessment for resective epilepsy surgery. The degree of the impact will be dependent on how many people decide to undergo assessment for resective epilepsy once referred because the assessment for resective epilepsy surgery is resource intensive. The committee noted that even if more people are referred for assessment for resective epilepsy the proportion of people who are eligible for surgery and proceed to resective epilepsy surgery is unlikely to change substantially.

1.4.5. Other factors the committee took into account

The committee noted that the evidence was only in people who were already drug-resistant, and that the comparator in these studies was medical management. This initially suggests a certain bias in the studies over and above that evaluated in the risk of bias assessments, resulting from the samples being made up of people who would be predisposed to do badly on the comparator treatment. This would naturally increase the likelihood for the intervention to appear superior. However, it was also realised that there are only two established ways to treat epilepsy – drugs or surgery. If the drugs work well, then surgery would probably not be contemplated, but if the drugs are ineffective, then surgery is a viable option. Hence it is correct that surgery should be tested for efficacy in the population where the drugs don’t achieve their aim.

1.4.6. Recommendations supported by this evidence review

This evidence review supports recommendations 8.2.1 – 8.2.4.

A.1. Review protocol for surgery referral criteria

Download PDF (173K)

A.2. Review protocol for surgical interventions

Download PDF (229K)

A.3. Health economic review protocol

Download PDF (171K)

B.1. Surgical interventions

Download PDF (269K)

B.2. Referral for surgery

Download PDF (270K)

G.1. Referral to surgery

Download PDF (158K)

G.2. Surgical interventions

Download PDF (138K)

I.1. Surgery versus waitlist-control/medical treatment

Download PDF (265K)