Introduction
Epilepsy is a chronic neurological disorder characterized by recurrent seizures. In Canada, the prevalence of self-reported epilepsy ranges from 5.2 to 5.6 cases per 1000 population.Reference Tellez-Zenteno, Pondal-Sordo, Matijevic and Wiebe 1 Every year, approximately 15,500 new patients are diagnosed with this condition.Reference Theodore, Spencer, Wiebe, Langfitt, Ali and Shafer 2 For most individuals affected by epilepsy, seizures can be brought under control by drug therapy; however, up to 20-30% of patients do not respond to medication and surgical resection of the epileptic focus may be considered.Reference Burch, Hinde, Palmer, Beyer, Minton and Marson 3 The initial stage of the workup for surgery usually involves a series of tests to isolate the brain region responsible for the occurrence of seizures. Standard assessment consists of history and physical examination, prolonged scalp video-electroencephalogram (EEG) in an epilepsy monitoring unit, magnetic resonance imaging (MRI) of the brain, and neuropsychological testing. When there is a clear lesion and the video-EEG results coincide with the MRI lesion, patients will undergo surgical resection of the epileptogenic focus. But, in those cases where the information obtained is not concordant or does not provide an accurate localization, intracranial placement of electrodes and subsequent video-EEG (intracranial EEG) may be indicated. Currently, noninvasive studies provide information to guide the placement of intracranial EEG electrodes. If the seizure focus is localized, surgery is considered. Precise presurgical localization of the seizure focus is essential to achieving good surgical outcomes.
Despite the long-standing application of fludeoxyglucose F 18 (18F-FDG) positron emission tomography (PET) in the presurgical evaluation of patients with medically intractable epilepsy, the role of this technology continues to be refined with usage differing among providers and institutions. PET has the unique capability of imaging cerebral metabolism, whereas EEG measures electrical activity and MRI depicts only gross anatomic alterations associated with epilepsy. Each test is of clinical value and can provide information that can be used for all levels of surgical decision-making. Several reports in the past have indicated that PET is safe and may benefit a subset of patients undergoing surgery,Reference Dussault, Nguyen and Rachet 4 - 6 whereas another report concluded that there is a lack of evidence on the effectiveness and cost-effectiveness of imaging techniques (including PET) in the presurgical workup to inform clinical practice.Reference Whiting, Gupta, Burch, Mota, Wright and Marson 7 As a result, the purpose of the present study was to systematically review the literature to assess the diagnostic accuracy and clinical utility of PET in the presurgical evaluation of adult and pediatric patients with medically intractable epilepsy.
Methods
The literature was searched using MEDLINE (1946 to September week 4 2013) and EMBASE (1974 to 2013 week 29) databases in OVID. The search strategy combined disease-specific terms (exp epilepsy/or epilep$.ti,ab.) with intervention-specific terms (exp tomography, emission computed/ or pet or positron emission tomograph$ or positron-emission),ti, ab.). In addition, annual meetings of the American Epilepsy Society were searched up to September 2013 for other relevant abstracts. Likewise, the Canadian Medical Association Infobase, the National Guidelines Clearinghouse, and the Cochrane Database of Systematic Reviews were searched up to September 2013 for existing evidence-based practice guidelines. Relevant articles and abstracts were selected and reviewed by two reviewers, and the reference lists from these were searched for additional studies, as were the reference lists from relevant review articles.
The following criteria were used to include studies: (1) fully published reports or abstracts of systematic reviews, randomized controlled trials, and prospective or retrospective studies that evaluated the use of 18F-FDG PET in medically intractable epilepsy; (2) studies that included ≥12 patients of any age; (3) reported on at least one of the following outcomes: diagnostic accuracy (sensitivity, specificity, positive predictive value [PPV], negative predictive value [NPV]), surgical management impact, or patient outcome impact; and (4) studies that used a suitable reference standard (intracranial EEG, surgical eligibility, good surgical outcome [Engel class I, II, or III]) when appropriate. The exclusion criteria were (1) studies of non–18F-FDG PET; (2) nonsystematic reviews, letters, editorials, individual case reports, historical articles, or commentaries; and (3) reports published in a language other than English. An assessment of study quality was performed for all fully published reports by one reviewer.
Results
No existing systematic reviews or evidence-based guidelines were found that specifically evaluated the use of 18F-FDG PET against a suitable reference standard. In addition, there were no randomized controlled trials comparing the diagnostic accuracy and clinical utility of 18F-FDG PET with intracranial EEG. However, 37 retrospective studiesReference Heinz, Ferris, Lee, Radtke, Crain and Hoffman 8 - Reference Snead, Chen, Mitchell, Kongelbeck, Raffel and Gilles 43 , Reference Chassoux, Rodrigo, Semah, Beuvon, Landre and Devaux 44 and three prospective studiesReference Knowlton, Elgavish, Bartolucci, Ojha, Limdi and Blount 45 - Reference Theodore, Sato, Kufta, Gaillard and Kelley 47 were identified to be relevant to this systematic review (Figure 1). Six of these studies were reported solely in abstract form,Reference Sucak, Kapucu, Capraz, Akdemir, Kurt and Hirfanoglu 17 , Reference Dickson, Rathore, Ell and Duncan 24 , Reference Popescu, Sara, Mai, Minniti, Zanni and Rossetti 25 , Reference Eddeine and Chung 34 , Reference Piantino and Hussein 38 , Reference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 whereas two studiesReference Rubi, Setoain, Donaire, Bargallo, Sanmarti and Carreno 30 , Reference Desai, Bekelis, Thadani, Roberts, Jobst and Duhaime 33 had both the full publication and the abstract. Because of the heterogeneity of the studies in the patient population, study design, outcome measurements, and methods of PET interpretation, the results of the studies included in the systematic review could not be pooled. Instead, a qualitative analysis of the results was performed.
Figure 1 Literature flow diagram.
Study Design and Quality
For the fully published reports, study quality was assessed using the QUADAS-2 tool (Table 1). Abstracts were not assessed because of limited reporting of study information. The overall quality varied among the studies, but the large majority were judged to have a low risk of bias. The most common concern was the influence of PET results on the interpretation of the reference standard. That is, localization with intracranial EEG, decision to perform surgery, and classification of surgical outcomes were often not blinded to PET findings. Furthermore, some studies excluded patients with MRI abnormalities (i.e. structural lesions),Reference Kim, Lee, Lee, Chung, Chung and Lee 13 , Reference Lee, Lee, Kim, Hong, Lee and Chung 16 , Reference Yun, Lee, Lee, Kim, Jeong and Chung 19 , Reference DellaBadia, Bell, Keyes, Mathews and Glazier 21 , Reference van Huffelen, van Isselt, van Veelen, van Rijk, van Bentum and Dive 37 , Reference Widjaja, Shammas, Vali, Otsubo, Ochi and Snead 46 , Reference Theodore, Sato, Kufta, Gaillard and Kelley 47 incomplete tests or short follow-up,Reference Seo, Holland, Rose, Rozhkov, Fujiwara and Byars 39 lost to follow-up,Reference Yun, Lee, Lee, Kim, Jeong and Chung 19 or a definite extratemporal seizure origin.Reference Uijl, Leijten, Arends, Parra, Van Huffelen and Moons 26
Table 1 QUADAS-2 assessment of study quality
H, high risk; L, low risk; QUADAS, Quality Assessment of Diagnostic Accuracy Studies; U, unclear risk.
Diagnostic Accuracy
Comparison with Intracranial EEG
There were eight retrospective studies identified that investigated the localization of seizure foci with PET compared with intracranial EEG in adult patients.Reference Debets, van Veelen, Maquet, van Huffelen, van Emde Boas and Sadzot 31 - Reference van Huffelen, van Isselt, van Veelen, van Rijk, van Bentum and Dive 37 , Reference Widjaja, Shammas, Vali, Otsubo, Ochi and Snead 46 These studies included patients with temporal and/or extratemporal lobe epilepsy. One of the studiesReference Debets, van Veelen, Maquet, van Huffelen, van Emde Boas and Sadzot 31 reported positive correlation between PET and intracranial EEG for both localization (59%) and lateralization (18%) of onset. That is, using intracranial EEG as the reference standard, PET correctly identified the epileptogenic lobe in 59% of the patients and the epileptogenic side, but not the lobe in 18% of the patients. Another study reported a sensitivity of 77% for lateralization only.Reference Tatlidil, Luther, West, Jadvar and Kingman 36 Overall, the sensitivity at which PET hypometabolism agreed with seizure onset localized by intracranial EEG ranged from 56% to 90% (weighted mean=71%). Among studies that included only temporal lobe epilepsy patients,Reference Delbeke, Lawrence, Abou-Khalil, Blumenkopf and Kessler 32 , Reference Eddeine and Chung 34 , Reference Theodore, Sato, Kufta, Gaillard and Kelley 47 the sensitivity of PET ranged from 63% to 90% (weighted mean=74%) (Table 2).
Table 2 Comparison between 18F-FDG PET and intracranial EEG in the localization of seizure foci (adult)
BZ, benzodiazepine; NR, not reported; Q, quantitative; SPECT, single photon emission computed tomography; SQ, semiquantitative; TLE, temporal lobe epilepsy.
* Localization sensitivity=number of patients in whom PET localized the seizure focus that was concordant with intracranial EEG/total number of patients in whom the seizure focus was localized with intracranial EEG.
† Lateralization sensitivity=number of patients in whom PET lateralized (but not localized) the seizure focus that was concordant with intracranial EEG/total number of patients in whom the seizure focus was localized with intracranial EEG.
In pediatric patients with intractable epilepsy of both temporal and extratemporal origins, four primary studies were identified that compared PET with intracranial EEG in the localization of seizure foci.Reference Kumar, Juhasz, Asano, Sood, Muzik and Chugani 23 , Reference Piantino and Hussein 38 - Reference Seo, Noh, Lee, Kim, Lee and Kim 40 In one study,Reference Kumar, Juhasz, Asano, Sood, Muzik and Chugani 23 the results for two methods of PET interpretation—visual analysis (V) and statistical parametric mapping (SPM)—were reported. SPM using a threshold of p<0.001 provided a sensitivity of 86% when measured against intracranial EEG. The sensitivity decreased to 60% after using a stricter threshold of p<0.05. In comparison, the sensitivity for V was 74%. Another studyReference Seo, Holland, Rose, Rozhkov, Fujiwara and Byars 39 reported lobar concordance between PET and intracranial EEG in 21% of the patients and hemispheric but not lobar concordance in 50% of the patients. In general, the sensitivity of PET localization with respect to intracranial EEG varied from 21% to 86% (weighted mean=68%) across the four studies (Table 3).
Table 3 Comparison between 18F-FDG PET and intracranial EEG in the localization of seizure foci (pediatric)
MSI: Magnetic source imaging; NR: not reported.
* Localization sensitivity=number of patients in whom PET localized the seizure focus that was concordant with intracranial EEG/total number of patients in whom the seizure focus was localized with intracranial EEG.
† Lateralization sensitivity=number of patients in whom PET lateralized (but not localized) the seizure focus that was concordant with intracranial EEG/total number of patients in whom the seizure focus was localized with intracranial EEG.
‡ Values for numerator and denominator unavailable.
With Respect to Surgical Decision Making
Four retrospective studies examined the contribution of PET to surgical decision-making for adult patients with medically intractable epilepsy.Reference DellaBadia, Bell, Keyes, Mathews and Glazier 21 , Reference Mastin, Drane, Gilmore, Helveston, Quisling and Roper 22 , Reference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 , Reference Struck, Hall, Floberg, Perlman and Dulli 42 Two of these evaluated only patients with temporal lobe epilepsy,Reference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 , Reference Struck, Hall, Floberg, Perlman and Dulli 42 whereas in the other studies, patients with temporal and extratemporal lobe epilepsy were included.Reference DellaBadia, Bell, Keyes, Mathews and Glazier 21 , Reference Mastin, Drane, Gilmore, Helveston, Quisling and Roper 22 Two studiesReference DellaBadia, Bell, Keyes, Mathews and Glazier 21 , Reference Struck, Hall, Floberg, Perlman and Dulli 42 evaluated the predictive utility of PET on surgical eligibility. PET could accurately predict surgical candidacy in 68% (PPV) of the patients, which was equivalent to that of MRI and EEG. However, PET was the most sensitive (86%) and had the highest proportion of true-positive and true-negative tests (72%), whereas the sensitivity and proportion of true-positive and true-negative tests were 66% and 67%, respectively, for both MRI and EEG.Reference DellaBadia, Bell, Keyes, Mathews and Glazier 21 The second study also reported a sensitivity of 86% for PET, which was higher than that of EEG (82%) but lower than MRI (90%). Additionally, multivariate analysis revealed that PET hypometabolism was a significant predictor of postoperative outcome (p=0.02).Reference Struck, Hall, Floberg, Perlman and Dulli 42 Site of surgery was used as the reference standard in the other two studies.Reference Mastin, Drane, Gilmore, Helveston, Quisling and Roper 22 , Reference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 The abstract by Khan et alReference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 reported that 59% of the patients had either lateralizing or localizing PET findings corresponding to the resected seizure focus. The second studyReference Mastin, Drane, Gilmore, Helveston, Quisling and Roper 22 reported a similar sensitivity of 60% as well as a PPV of 83% (Table 4). In most of the studies, consensus agreement based on all available clinical and diagnostic information was used to determine surgical candidacy or surgical sites.
Table 4 Diagnostic accuracy of 18F-FDG PET with respect to surgery
ILAE, International League Against Epilepsy; NR, not reported; TLE, temporal lobe epilepsy.
* Sensitivity=number of surgical candidates with positive PET findings/total number of patients eligible for surgery (positive PET finding is defined as imaging abnormality in the area of surgical resection or conclusive evidence consistent with the final consensus decision regarding surgical candidacy).
† Specificity=number of patients considered ineligible for surgery on the basis of PET findings/total number of patients ineligible for surgery.
‡ PPV=proportion of patients accurately predicted to be eligible for surgery.
¶ NPV=proportion of patients accurately predicted not to be eligible for surgery.
|| Values for numerator and denominator unavailable.
One retrospective study evaluated the diagnostic performance of PET with respect to site of surgical resection in children with intractable epilepsy. Kumar et alReference Kumar, Juhasz, Asano, Sood, Muzik and Chugani 23 compared the results between V and SPM. The reported sensitivity from that study was 62% for V and 71% for SPM using a threshold of p<0.001 (35% with a stricter threshold of p<0.05). The specificity (V=89%; SPMp<0.001=86% to SPMp<0.05=98%) and PPVs (V=82%; SPMp<0.001=79% to SPMp<0.05=95%) were higher for both methods of analysis (Table 4). Resection margins were ultimately decided by intracranial EEG.
Patients with Good Surgical Outcome
In adult patients, a total of 13 primary studies used good surgical outcome to estimate the diagnostic accuracy of PET. Of these studies, 12 were retrospectiveReference Heinz, Ferris, Lee, Radtke, Crain and Hoffman 8 - Reference Yun, Lee, Lee, Kim, Jeong and Chung 19 and one was part of a prospective observational study.Reference Knowlton, Elgavish, Bartolucci, Ojha, Limdi and Blount 45 Good surgical outcome was considered in patients with Engel class I, II, or III. When outcomes were not reported by Engel’s classification, seizure-free or significantly improved (<10 seizures per year and at least a 90% reduction in seizures from the preoperative year) was considered good surgical outcome. Two studies that included only patients with temporal lobe epilepsyReference Kassem, El Shiekh, Wafaie, Abdelfattah, Farghaly and Afifi 11 , Reference Sucak, Kapucu, Capraz, Akdemir, Kurt and Hirfanoglu 17 reported separate sensitivity values for the magnetic resonance–positive and magnetic resonance–negative subgroups. The results were similar between the studies for the magnetic resonance–positive (88% and 89%) and magnetic resonance–negative (80% and 81%) patients. Overall, the proportion of patients in whom PET correctly localized a seizure focus and had a good surgical outcome ranged from 36% to 89% (weighted mean=68%). This range improved to 71% to 89% (weighted mean=86%) when only considering patients with temporal lobe epilepsy.Reference Heinz, Ferris, Lee, Radtke, Crain and Hoffman 8 , Reference Kassem, El Shiekh, Wafaie, Abdelfattah, Farghaly and Afifi 11 , Reference Sucak, Kapucu, Capraz, Akdemir, Kurt and Hirfanoglu 17 In contrast, the sensitivity of PET ranged from 36% to 66% (weighted mean=50%) in extratemporal lobe epilepsy patients only.Reference Kim, Lee, Yun, Kim, Lee and Chung 12 - Reference Lee, Lee, Lee, Park, Kim and Lee 15 PET was able to further lateralize the seizure focus in 13% to 29% (weighted mean=17%) of patients with a good surgical outcome.Reference Hong, Lee, Kim, Lee and Chung 9 , Reference Kim, Lee, Yun, Kim, Lee and Chung 12 - Reference Lee, Lee, Lee, Park, Kim and Lee 15 In one study,Reference Won, Chang, Cheon, Kim, Lee and Han 18 only the sensitivity for correct lateralization was reported (86%); therefore, it is not clear as to whether this is separate from or considered with localization. Lateralizing information gained from PET imaging is useful for planning an invasive study. In the only prospective study,Reference Knowlton, Elgavish, Bartolucci, Ojha, Limdi and Blount 45 the authors reported a sensitivity of 59%, a specificity of 79%, a PPV of 83%, and a NPV of 54% for Engel class I outcome (Table 5). These diagnostic values were similar to magnetic source imaging (56% sensitivity, 79% specificity, 82% PPV, and 52% NPV).
Table 5 Diagnostic accuracy of 18F-FDG PET in patients with good surgical outcome (adult)
FLE, frontal lobe epilepsy; ICEEG: intracranial EEG; NR, not reported; OLE, occipital lobe epilepsy; SPECT, single photon emission computed tomography.
* Localization sensitivity=number of patients in whom PET localized the seizure focus that was concordant with the surgical site and achieved good surgical outcome/total number of patients with good surgical outcome.
† Lateralization sensitivity=number of patients in whom PET lateralized (but not localized) the seizure focus that was concordant with the surgical site and achieved good surgical outcome/total number of patients with good surgical outcome.
‡ Specificity=number of patients with negative PET findings and did not achieve good surgical outcome/total number of patients who did not achieve good surgical outcome (negative PET finding is defined as normal or multilobar pattern in both hemispheres).
¶ PPV=proportion of PET positive patients accurately predicted to achieve good surgical outcome (positive PET finding is defined as imaging abnormality in the area of surgical resection or conclusive evidence consistent with the final consensus decision regarding surgical candidacy).
|| NPV=proportion of PET negative patients accurately predicted to not achieve good surgical outcome (negative PET finding is defined as normal or multilobar pattern in both hemispheres).
# Significantly improved is defined <10 seizures per year and ≥90% reduction in seizures from the preoperative year.
** Values for numerator and denominator unavailable.
In Engel class I pediatric patients, one prospective studyReference Knowlton, Elgavish, Bartolucci, Ojha, Limdi and Blount 45 evaluated the sensitivity (65%), specificity (94%), PPV (68%), and NPV (94%) of PET relative to lobar localization. The corresponding values for magnetoencephalography were 85%, 99%, 94%, and 97%, respectively. However, if one or both of the two tests were concordant with cortical resection, the sensitivity increased to 95%. In one retrospective study,Reference Khan, Appel, Dustin, Reeves-Tyer, Bagic and Martinez 41 PET showed a localizing sensitivity of 73% for temporal lesions and 63% for extratemporal lesions. The corresponding lateralizing sensitivities for temporal and extratemporal cases were 23% and 5%, respectively (Table 6).
Table 6 Diagnostic accuracy of 18F-FDG PET in patients with good surgical outcome (pediatric)
MEG, magnetoencephalography; NR, not reported; SISCOM, subtraction of ictal and interictal single photon emission computed tomography coregistered to MRI.
* Localization sensitivity=number of patients in whom PET localized the seizure focus that was concordant with the surgical site and achieved good surgical outcome/total number of patients with good surgical outcome.
† Lateralization sensitivity=number of patients in whom PET lateralized (but not localized) the seizure focus that was concordant with the surgical site and achieved good surgical outcome/total number of patients with good surgical outcome.
‡ Specificity=number of patients with negative PET findings and did not achieve good surgical outcome/total number of patients who did not achieve good surgical outcome (negative PET finding is defined as normal or multilobar pattern in both hemispheres).
¶ PPV=proportion of PET positive patients accurately predicted to achieve good surgical outcome (positive PET finding is defined as imaging abnormality in the area of surgical resection or conclusive evidence consistent with the final consensus decision regarding surgical candidacy).
|| NPV=proportion of PET negative patients accurately predicted to not achieve good surgical outcome (negative PET finding is defined as normal or multilobar pattern in both hemispheres).
# Values for numerator and denominator unavailable.
Impact on Patient Management
18F-FDG PET
The evidence demonstrating the impact of PET on clinical management in adult patients came from three retrospective studies. In the Uijl et al study,Reference Uijl, Leijten, Arends, Parra, Van Huffelen and Moons 26 the impact of PET was assessed by comparing documented decisions regarding surgical candidacy before and after PET findings. The initial decision concerning whether to perform temporal lobe epilepsy surgery was based on MRI and video-EEG findings, and PET results led clinicians to change their decision in 71% (78 of 110) of the patients who underwent PET (of these 78 patients, 28 avoided surgery, 48 were considered for surgery [62% had Engel class I surgical outcome], and two were requested for intracranial monitoring [one was subsequently considered for surgery and had Engel class I surgical outcome, whereas surgery was ultimately not performed in the other). The abstract by Dickson et alReference Dickson, Rathore, Ell and Duncan 24 assessed the benefit of PET in the presurgical evaluation of 194 consecutive patients with medically refractory focal epilepsy. In this study, PET findings led directly to surgery in 6% of the cases, helped in planning intracranial EEG in 35% of the cases, and excluded 12% of the cases from additional evaluation. In another abstract by Popescu et al,Reference Popescu, Sara, Mai, Minniti, Zanni and Rossetti 25 a preliminary study was undertaken to study the role of V and SPM analysis of PET in patients with temporal and extratemporal epilepsy. Results from the study showed that both methods of analysis helped improve the guidance of intracranial electrodes placement in 48% of the patients and ruled out stereo-EEG in 21% of the patients (Table 7).
Table 7 Impact of 18F-FDG PET on patient management
NR, not reported; SQ, semiquantitative; TLE, temporal lobe epilepsy.
There were three retrospective studies that provided evidence of a change in clinical management in pediatric patients because of PET. One studyReference Chugani and Conti 27 investigated the effectiveness of PET in classifying symptomatic infantile spasms. With the benefit of PET, the number of cases classified as symptomatic increased from 30% to 96%. In other words, PET uncovered unifocal or multifocal metabolic abnormalities in 95% of the cryptogenic cases. In the study by Ollenberger et al,Reference Ollenberger, Byrne, Berlangieri, Rowe, Pathmaraj and Reutens 28 the role of PET in the diagnosis and management of children with refractory epilepsy was assessed from the clinician’s perspective. Three epileptologists completed the questionnaires in reference to 113 evaluable patients. For surgical candidates, PET scan results excluded surgery (major impact) in 39% of the patients and modified surgery (minor impact) in 19% of the patients. For medical therapy patients, PET resulted in surgery being excluded in 5% of the patients and management plan modified in 19% of the patients. The third studyReference Snead, Chen, Mitchell, Kongelbeck, Raffel and Gilles 43 compared children who received PET as part of epilepsy surgery evaluation (n=56) with those who did not (n=44). The authors reported that there was no significant difference between the two groups in terms of the number of children who underwent surgery, the type of procedure performed, the clinical outcome, or whether chronic invasive intracranial monitoring was needed. Of the 16 patients who had focal cortical resection or hemispherectomy, three avoided invasive monitoring because of localizing information provided by PET (Table 7).
18F-FDG PET/MRI coregistration
There were three primary studies that investigated the value of incorporating PET/MRI coregistration into the presurgical evaluation of patients with medically intractable epilepsy. The retrospective study by Salamon et alReference Salamon, Kung, Shaw, Koo, Koh and Wu 29 compared two cohorts of patients with cortical dysplasia (CD), one in which PET/MRI coregistration was a routine part of the presurgical evaluation (n=45) and the other without (n=38). Compared with the patients before the regular use of PET/MRI coregistration, the cohort with the benefit of this technique had 18% more patients receiving surgery, a higher proportion of patients with type I CD on histopathology (60% versus 24%; p=0.0009), and fewer patients undergoing intracranial electrode studies (2% vs 21%; p=0.0060). In this same cohort, surgical resection guided by PET/MRI coregistration and electrocorticography resulted in 82% of the patients achieving seizure freedom. In another retrospective study involving children with refractory epilepsy,Reference Rubi, Setoain, Donaire, Bargallo, Sanmarti and Carreno 30 PET/MRI coregistration guided the second MRI interpretation from nonlesional to subtle lesional in 42% of the cases. Similarly, PET/MRI coregistration was able to detect Taylor-type focal cortical dysplasia in patients with negative MRI and where a PET scan alone does not allow a conclusive diagnosis. Cortical resection guided by PET/MRI coregistration in addition to stereo-EEG led to 87% of patients achieving seizure freedom (Table 8).Reference Chassoux, Rodrigo, Semah, Beuvon, Landre and Devaux 44
Table 8 Impact of 18F-FDG PET/MRI coregistration on patient management
CD, cortical dysplasia.
Discussion
In patients with medically intractable epilepsy, the main goal of presurgical evaluation is to provide precise localization of the epileptogenic focus with the intention of optimally selecting surgical candidates who are likely to have a seizure-free outcome after resective surgery. To date, no single test alone has been sufficient for localizing the surgical site and evaluation is based on a consensus of all available diagnostic information. Numerous scenarios arise in which intracranial EEG is necessary to provide critical data for patient management. However, intracranial EEG is an invasive procedure and poses the risk (although low) of infection, hemorrhage, and cerebral edema.Reference Burneo, Steven, McLachlan and Parrent 48 Particularly in children, the hospital stay is lengthened because of the time required to obtain the ictal onset and functional mapping information. With modern advances in structural and functional imaging, the ability to provide accurate information without the need for intracranial EEG has become increasingly important. In many patients, intracranial EEG can be avoided when data from less-invasive studies are concordant in their lateralization and localization.
FDG PET has been known to indirectly localize the seizure focus by determining areas of decreased glucose metabolism. A previous meta-analysis reported a concordance value of 67% between PET and invasive EEG recording.Reference Willmann, Wennberg, May, Woermann and Pohlmann-Eden 49 This analysis included only patients with temporal lobe epilepsy and excluded pediatric patients. Still, data from this systematic review are consistent with their findings and showed a 56 to 90% agreement between PET hypometabolism and seizure onset localized by intracranial EEG among adults. Similar results were observed in pediatric patients except for one study that reported only 21% of patients in whom PET correctly localized the seizure focus when measured against intracranial EEG. However, PET was able to lateralize a further 71%. In the other studies, it was not possible to distinguish between localizing and lateralizing findings because this information is often hidden, not separated, or considered the same.
Despite the general acceptance of intracranial EEG as the gold standard for localizing the seizure onset, in clinical practice, the decision to proceed with surgery may come from a number of sources. Therefore, surgical candidacy or site of surgical resection was also considered as a reference standard for this review. Based on these studies, PET demonstrated significant influence on surgical decision making in adults, with moderate to high sensitivities and PPVs. In children, SPM analysis of PET performed similarly well in the identification of surgical resection areas.
The ultimate reference standard for successful localization is surgical outcome. In adults, the data showed high sensitivity (88% and 89%) for PET with respect to good surgical outcome when MRI is positive. Although the overall sensitivity of PET varied considerably across the studies, PET displayed moderate to high sensitivity in localizing the seizure focus among temporal lobe epilepsy patients (range, 71-89%). Again, this is in line with the previous meta-analysis which showed 86% of patients with good outcome had an ipsilateral PET finding (49). For MRI and EEG, the reported sensitivity (MRI 41-83%; EEG 36-81%) also varied greatly across the studies.Reference Heinz, Ferris, Lee, Radtke, Crain and Hoffman 8 - Reference Lee, Lee, Kim, Hong, Lee and Chung 16 , Reference Won, Chang, Cheon, Kim, Lee and Han 18 , Reference Yun, Lee, Lee, Kim, Jeong and Chung 19 Perhaps of greater importance is when PET results were combined with MRI or EEG, the sensitivity of detecting patients with good outcome increased by 8 to 23%.Reference Heinz, Ferris, Lee, Radtke, Crain and Hoffman 8 - Reference Hwang, Kim, Park, Han, Yu and Lee 10 , Reference Kim, Lee, Lee, Chung, Chung and Lee 13 In children, the addition of PET to magnetoencephalography increased the sensitivity to 95% and decreased the number of false-negative tests for seizure-free outcome.Reference Widjaja, Shammas, Vali, Otsubo, Ochi and Snead 46 Previous studies have suggested that 55 to 70% of patients undergoing temporal resection achieve a completely seizure-free state, whereas only 30 to 50% of patients undergoing extratemporal resection achieve seizure freedom.Reference Tonini, Beghi, Berg, Bogliun, Giordano and Newton 50 The results of the present study suggest that localization is greater in patients with temporal lobe epilepsy, who are more likely to benefit from surgical treatment than in patients with extratemporal lobe epilepsy. It appears that the heterogeneous clinical features of extratemporal (i.e. frontal, insular, occipital, and parietal) epilepsy make accurate localization more difficult. This is a critical issue in children in whom medically refractory extratemporal focal epilepsy is more common in surgical candidates than that of temporal origin. The reverse is true in adult epilepsy surgery candidates. PET findings have been shown to impact patient management by improving the guidance of intracranial electrodes placement, altering the decision to perform surgery or excluding patients from further evaluation.
Because of variable population characteristics (age, types of epilepsy), outcome measurements (inconsistent use of Engel’s classification system in reporting surgical outcome), and methods of PET interpretation (V, quantitative, semiquantitative, SPM) among the studies, a meta-analysis was not performed. Instead, a narrative synthesis of the results was presented. The majority of the available studies were retrospective studies with a greater proportion of the evidence in adult patients. This can lead to the introduction of selection bias because only patients proceeding to surgery can be included when surgical outcome was used as a reference standard. Additionally, many of the studies did not report on test specificity, but would be relevant in determining the ability of PET to exclude patients who are unlikely to be amenable to surgery. Although the ideal evidence for evaluating the clinical utility of PET derives from randomized controlled trials, their conduct in this area may not be feasible because of ethical issues.
Currently, FDG PET is widely accepted and recognized as a complementary technique in the presurgical assessment by most epilepsy centres around the world. The combination of imaging findings in relation to each other can enable more accurate localization for surgical resection. Thus, PET can be useful in this setting, particularly in temporal lobe epilepsy patients whose MRI is negative and/or have discordant localizing/lateralizing data from other diagnostic modalities.
Conclusion
The potential benefit of PET in the presurgical evaluation of patients with intractable epilepsy lies in its ability to provide data for localizing the seizure focus and to determine resectability. The evidence from this review proposes that PET is able to provide complementary information that can guide decision-making toward successful surgery. Nonetheless, there is a need for prospective studies to assess the use of PET/MRI and the advantages over standard PET studies.
Disclosures
JGB has received honoraria from and given talks for UCB Canada, EISAI, and Sunovion. The other authors have no disclosures to report.
