- Top of page
Purpose:[18F]-Fluoro-d-deoxyglucose positron emission tomography (FDG-PET) is an expensive, invasive, and not widely available technique used in the presurgical evaluation of temporal lobe epilepsy. We assessed its added value to the decision-making process in relation to other commonly used tests.
Methods: In a retrospective study of a large series of consecutive patients referred to the national Dutch epilepsy surgery program between 1996 and 2002, the contribution of FDG-PET, magnetic resonance imaging (MRI), and video-electroencephalogram (video-EEG) monitoring findings, alone or in combination, to the decision whether to perform surgery was investigated. The impact of FDG-PET was quantified by comparing documented decisions concerning surgery before and after FDG-PET results.
Results: Of 469 included patients, 110 (23%) underwent FDG-PET. In 78 of these patients (71%), FDG-PET findings led clinicians to change the decision they had made based on MRI and video-EEG monitoring findings. In 17% of all referred patients, the decision regarding surgical candidacy was based on FDG-PET findings. FDG-PET was most useful when previous MRI results were normal (p < 0.0001) or did not show unilateral temporal abnormalities (p < 0.0001), or when ictal EEG results were not consistent with MRI findings (p < 0.0001) or videotaped seizure semiology (p = 0.027). The positive and negative predictive values for MRI and video-EEG monitoring, which ranged from 0.48 to 0.67, were improved to 0.62 to 0.86 in combination with FDG-PET.
Conclusions: In patients referred for TLE surgery, FDG-PET findings can form the basis for deciding whether a patient is eligible for surgery, and especially when MRI or video-EEG monitoring are nonlocalizing.
[18F]-Fluoro-d-deoxyglucose positron emission tomography (FDG-PET) is used in the complex presurgical evaluation of patients with medically intractable temporal lobe epilepsy (TLE) (Debets et al., 1990, 1997; Engel, 1993; Juhasz and Chugani, 2003). However, because the technique is invasive and expensive, requiring the injection of radioactivity, and not widely available because cyclotron facilities are needed (Williamson and Jobst, 2000), it is important to know its diagnostic value compared with that of more routinely performed investigations such as magnetic resonance imaging (MRI) and video-electroencephalogram (video-EEG) seizure recordings.
Recent studies suggest that FDG-PET is indicated in patients with TLE if MRI does not localize the source in one temporal lobe (either because it is negative or shows bilateral abnormalities) or if ictal EEG findings show a unilateral temporal onset that is not consistent with MRI findings (Henry, 2001; O'Brien et al., 2001; Carne et al., 2004). Although several studies have investigated the contribution of FDG-PET to identifying the lobe of seizure onset in patients with TLE, in most studies FDG-PET was evaluated in isolation, without reference to existing MRI and video-EEG monitoring results (Uijl et al., 2005). Medical decisions are usually based on the results of several investigations and are hardly ever based on a single test result (Moons and Grobbee, 2002a). This is also true for the diagnostic work-up of patients regarding their eligibility for TLE surgery, in which FDG-PET is never performed first. The aim of this study was to determine the clinical or added value of FDG-PET on the decision-making process regarding TLE surgery in the setting of a tertiary referral center. We were particularly interested in determining the contribution of FDG-PET relative to that of MRI and video-EEG monitoring.
- Top of page
Between January 1996 and July 2002, 632 patients were referred to the national DCESP for evaluation for epilepsy surgery (Fig. 1). Of these, 142 were excluded because the epileptogenic focus was considered to be extratemporal. Twenty-one patients withdrew from the diagnostic work-up. Therefore, 469 patients were included in the present analysis, of whom 302 (232 + 70; 60%) underwent surgery. FDG-PET was performed in 110 patients (23%), 70 of whom (64%) were considered eligible for surgery as compared with 232 of 359 patients (65%) who did not undergo FDG-PET. One year after surgery, 64% of all operated patients were completely seizure-free without auras (Engel class 1A); this was the case for 60% of the patients who had undergone FDG-PET. This difference was not statistically significant. The mean follow-up after surgery was 4.2 years, which ranged from 1 to 10 years. Of all operated patients, 51% reached complete seizure freedom at last follow-up; this was the case in 48% of patients who had undergone FDG-PET. Again, this difference was not statistically significant.
On the basis of MRI and video-EEG findings, 231 patients were considered eligible for TLE surgery (Fig. 2); the taskforce nevertheless decided to perform FDG-PET in 16 of these patients (7%), in most cases because the MRI was not localizing or simply to confirm MRI findings. On the basis of FDG-PET findings, the decision to operate was changed, or intracranial monitoring was requested, in 4 of these patients (25%). Of the 68 patients considered ineligible for surgery on the basis of MRI and video-EEG findings, the taskforce reconsidered and decided to perform FDG-PET in 10 patients (15%). One patient was subsequently considered eligible for surgery; however, this patient did not become seizure-free after surgery. Most patients who underwent FDG-PET had inconclusive results after MRI and video-EEG monitoring. Of the 84 of 170 patients with inconclusive results who underwent FDG-PET, FDG-PET led to a final decision in 72 patients (Fig. 2: 47 + 25; 86%). Compared with the rest of the cohort, these 72 patients more often had normal MRI findings (p < 0.0001), less often unilateral temporal abnormalities on MRI (p < 0.0001), MRI and ictal EEG were less often concordant (p < 0.0001), and videotaped semiology and ictal EEG were less often concordant (p = 0.027) (Table 1). The outcome after surgery of the operated patients with inconclusive results after MRI and video-EEG monitoring who underwent FDG-PET (63% seizure free) was comparable to that of all operated patients (64% seizure free). In total, FDG-PET findings were conclusive regarding surgical candidacy in 78 (Fig. 2; 4 + 2 + 72) of 469 patients (17%) referred for TLE surgery or in 78 of 110 patients (71%) investigated with FDG-PET. With the exception of interictal video-EEG, all included tests (MRI, video-EEG monitoring and FDG-PET findings) were associated with the final consensus decision regarding surgical candidacy (Table 2). However, in isolation, none of the tests showed good prediction or discrimination for this decision.
Table 1. Characteristics of the patient group with indecisive results after MRI and video EEG in whom FDG-PET results forced a decision compared to the rest of the cohort
| ||Decision forced by FDG-PET|
|Yes (N = 72)||No (N = 397)|
|Age (mean ± standard deviation)||33 ± 12||31 ± 13|
|Male sex||35 (0.49)||195 (0.49)|
|MRI normal||28 (0.39)||17 (0.04) **|
|MRI unilateral temporal||34 (0.47)||276 (0.70) **|
|MRI bilateral temporal|| 3 (0.04)||21 (0.05)|
|Interictal video-EEG unilateral temporal||21 (0.29)||129 (0.33)|
|Videotaped semiology temporal||30 (0.42)||206 (0.52)|
|Ictal video-EEG unilateral temporal||40 (0.56)||255 (0.64)|
|Concordance MRI—interictal video-EEG||23 (0.32)||142 (0.36)|
|Concordance MRI—videotaped semiology||13 (0.18)||106 (0.27)|
|Concordance interictal video-EEG—videotaped semiology||17 (0.24)||109 (0.28)|
|Concordance MRI—ictal video-EEG||14 (0.19)||179 (0.45) **|
|Concordance interictal video-EEG—ictal video-EEG|| 9 (0.13)||69 (0.17)|
|Concordance videotaped semiology—ictal video-EEG||11 (0.15)||105 (0.26) *|
Table 2. Two by two tables and decision predictive values (95% confidence intervals) of FDG-PET, MRI and video-EEG monitoring with the decision for or against surgery after the FDG-PET scan
| ||Decision after FDG-PET:||Positive decision predictive valued||Negative decision predictive valuee|
|Surgery N= 60||No surgerybN= 50||Total N= 110||p-valuec|
|FDG-PET|| ||0.009|| |
| Unilateral temporal||41||22||63|| ||0.65||0.60|
| Inconclusivea/normal||19||28||47|| ||(0.53–0.77)||(0.45–0.72)|
|MRI|| ||0.009|| |
| Unilateral temporal||35||17||52|| ||0.67||0.57|
| Inconclusivea/normal||25||33||58|| ||(0.54–0.78)||(0.44–0.69)|
|Video-EEG interictal|| ||NS|| |
| Unilateral temporal||22||15||37|| ||0.59||0.48|
| Inconclusivea||38||35||73|| ||(0.43–0.74)||(0.37–0.59)|
|Video-EEG semiology|| ||0.033|| |
| Definitely temporal||32||17||49|| ||0.65||0.54|
| Not localizing||28||33||61|| ||(0.51–0.77)||(0.42–0.66)|
|Video-EEG ictal|| ||0.035|| |
| Unilateral temporal||39||23||62|| ||0.63||0.56|
| Inconclusivea||21||27||48|| ||(0.50–0.74)||(0.42–0.69)|
Table 3 shows the contribution of FDG-PET combined with MRI and ictal EEG to the decision whether to perform surgery. The first row of each combination can be seen as ‘the combined positive test result’ and therefore reflects the positive decision predictive value (PDPV) of that test combination. The last row can be seen as ‘the combined negative test result’, and therefore as the negative decision predictive value (NDPV) of that test combination. These values can be compared to the PDPV and NDPV of each test in isolation (Table 2). Addition of FDG-PET improved the PDPV and NDPV of MRI from 0.67 and 0.57 to 0.77 and 0.68, respectively, and the PDPV and NDPV of ictal EEG from 0.63 and 0.56 to 0.80 and 0.67, respectively. The PDPV and NDPV of long-term interictal EEG increased from 0.59 and 0.48 to 0.71 and 0.62, and the PDPV and NDPV of seizure semiology increased from 0.65 and 0.54 to 0.86 and 0.70, respectively (data not shown).
Table 3. FDG-PET in combination with MRI and ictal EEG in relation to the decision for or against TLE surgery
|MRI||FDG-PET||Decision after PET||Total|
|Unilateral temporal||Concordant unilateral temporal||26 (0.77c)|| 8 (0.23)||34|
|Unilateral temporal||Inconclusivea/normal||9 (0.56)|| 7 (0.44)||16|
|Unilateral temporal||Discordant||0 || 2 || 2|
|Inconclusivea/normal||Unilateral temporal|| 15 (0.56) || 12 (0.44) ||27|
|Inconclusivea/normal||Inconclusivea/normal|| 10 (0.32) ||21 (0.68d)||31|
|Ictal EEG||FDG-PET|| |
|Unilateral temporal||Concordant unilateral temporal||28 (0.80c)||7 (0.20)||35|
|Unilateral temporal||Inconclusivea/normal|| 11 (0.48) ||12 (0.52) ||23|
|Unilateral temporal||Discordant||0 ||4 ||4|
|Inconclusivea||Unilateral temporal|| 13 (0.54) ||11 (0.46) ||24|
|Inconclusivea||Inconclusivea/normal|| 8 (0.33) ||16 (0.67d)||24|
FDG-PET findings were discordant with MRI or video-EEG findings in nine patients, one of whom was still considered eligible for surgery. This patient did not become seizure-free.
Table 4 shows the results after imputation of the decisions expected to have been made if the patients who had not undergone FDG-PET had undergone FDG-PET. The proportion of decisions regarding surgical candidacy hardly changed after imputation in the patients who were considered eligible for surgery on the basis of MRI or video-EEG findings and in the patients with inconclusive results before FDG-PET. This indicates that the observed results (Fig. 2) for these patient groups were unlikely to have been biased. Unfortunately, imputation was methodologically impossible for the patients considered ineligible for surgery on the basis of MRI and video-EEG findings (the second arm in Fig. 2) as there were only ten cases in this group, of whom only one was considered eligible for surgery after FDG-PET.
Table 4. Sensitivity analysis: observed decisions for or against surgery and estimated decisions after imputation of FDG-PET results
|Decision before FDG-PET||Decision after FDG-PET||Observed (see Fig. 2)||After imputation|
|N||Cases||Fraction||95% CI||N||Cases||Fraction||95% CI|
|No surgery|| || 3||0.19||0.07–0.43|| || 32||0.14||0.11–0.23|
|No surgery||Surgery||10|| 1||0.10||0.02–0.40||NP|| |
|No surgery|| || 8||0.80||0.49–0.94|| |
|No surgery|| ||25||0.30||0.21–0.40|| || 44||0.26||0.20–0.33|
- Top of page
In 71% of the TLE patients who underwent FDG-PET, the FDG-PET results influenced the decision whether or not temporal lobe surgery could be performed, and in 17% of all referrals for TLE surgery, FDG-PET findings were conclusive regarding surgical candidacy. One year after surgery, 64% of operated patients were seizure free (Engel class 1A). These findings indicate that FDG-PET has important added value for the decision-making process regarding TLE surgery in a tertiary referral setting. This is supported by the increased positive and negative decision predictive values of MRI and video-EEG monitoring in combination with FDG-PET. FDG-PET seemed especially valuable when MRI findings were normal, or when ictal EEG and MRI findings were not concordant, which is in line with indications for FDG-PET in the literature (Henry, 2001; O'Brien et al., 2001; Carne et al., 2004). Nevertheless, only 84 of 170 patients (49%) with indecisive results after MRI and video-EEG underwent FDG-PET. One can only guess why FDG-PET was not performed in the other patients. In The Netherlands, PET has always been available for epilepsy surgery purposes (although at first carried out in Liège, Belgium) and because most people have full medical insurance, financial considerations were unlikely to have had a role.
Although the role of FDG-PET in TLE surgery has been studied before, most studies have assessed FDG-PET as a single diagnostic test only or in relation to seizure outcome in operated patients only (the prognostic value of FDG-PET) (Uijl et al., 2005). DellaBadia et al. did address the contribution of a combination of sleep-deprived EEG, MRI, and FDG-PET to the decision-making process (DellaBadia et al., 2002). They found that FDG-PET was the most sensitive test when used in isolation. The positive predictive value of the combination of any two tests (with or without FDG-PET) was higher than that of FDG-PET in isolation. However, DellaBadia et al. investigated fewer patients (69 vs. 110 in our study). Ollenberger et al. also showed that FDG-PET had an impact on the clinical management of children referred for epilepsy surgery (Ollenberger et al., 2005), based on clinicians' personal point of view. They recommended that all children with epilepsy should undergo FDG-PET. However, most of the children in Ollenberger et al.'s study had extratemporal epileptic foci.
Some methodological limitations of our study need to be addressed. First, our study outcome was the final consensus decision on operability reached by our national taskforce. Although this is the best alternative in the absence of a formal reference standard (Swets, 1988; Moons and Grobbee, 2002a, 2002b; Bossuyt et al., 2003) and the outcome one year after surgery (64% seizure free) was comparable to that reported in the literature (Engel et al., 1998, 2003), there is no way to know whether patients were inappropriately rejected for surgery. Formally, only a randomized design (to operate or not) in patients referred for TLE surgery could settle this issue—which would in our view be unethical.
Second, the consensus decision of the multidisciplinary taskforce was based on all available information, including the results of FDG-PET under investigation. This might have introduced incorporation bias, which could have led to overestimation of the accuracy measures in Tables 2 and 3 (Ransohoff and Feinstein, 1978; Swets, 1988; Moons and Grobbee, 2002b). However, as we had systematically documented all intermediate consensus decisions before and after FDG-PET, we could study the change in decision-making due to the FDG-PET results, bypassing this incorporation bias.
Third, in 75% of FDG-PET investigations a flumazenil PET (FMZ-PET) was performed along with FDG-PET as part of scientific research (Debets et al., 1997). In these patients, the consensus decision after FDG-PET might have been influenced by the results of the FMZ-PET investigation. However, FMZ-PET results were similar or less informative than the FDG-PET results in most patients (90%), which is in agreement with earlier studies showing that FMZ-PET is not superior to FDG-PET in detecting the ictal onset zone (Debets et al., 1997; Ryvlin et al., 1998).
Lastly, since the results of MRI, video-EEG, and PET studies were reduced to a few variables, some diagnostic information may have been missed or simplified. The choice of test result categories, however, was based on the literature, clinical practice, and considerations of objectivity and reproducibility.
We conclude that FDG-PET has added value to clinical decision-making in patients referred for TLE surgery. FDG-PET seems especially valuable when MRI and video-EEG monitoring are unable to localize the epileptic focus. FDG-PET findings influenced clinical decision-making in 71% of the patients investigated with FDG-PET and were conclusive regarding surgical eligibility in 17% of patients referred for TLE surgery.
- Top of page
The authors thank Anja J. Couperus, EEG technician, for her help in coding the data from video-EEG monitoring into our research database. PET scans were performed in Liège, Belgium (B. Sadzot, MD, PhD) and at the Free University Medical Center of Amsterdam, the Netherlands (A.A. Lammertsma, PhD).
This study was made possible by grants from the National Board of Academic Hospitals and the Health Care Insurance Board (VAZ/CVZ grant 001134) and the Dutch National Epilepsy Foundation (NEF grant 04-05).