RESUMO
OBJECTIVE: To review the effectiveness and/or accuracy, cost-effectiveness, and predictive value of neuroimaging of the cerebral cortex to visualise the seizure focus in people with refractory epilepsy being considered for surgery. DATA SOURCES: Electronic databases, Internet searches, hand searching and consultation with experts. METHODS: A systematic review was undertaken according to published guidelines. Results of diagnostic accuracy studies were analysed according to the imaging test evaluated. For each study the proportion of patients who were correctly localised, not localised, partially localised or incorrectly localised by the index test was calculated. Due to the heterogeneity present between studies, statistical pooling was not performed. Instead, a narrative synthesis of results is presented. For studies using multivariate analysis to look at the association of neuroimaging findings and outcome following surgery, all factors considered in the analyses were presented. Studies were grouped according to the neuroimaging technique investigated and the findings discussed with reference to possible sources of heterogeneity between studies. RESULTS: No randomised controlled trials (RCTs) were identified, with the majority of studies evaluating the diagnostic accuracy of various imaging techniques in the localisation of epileptic seizure foci. There was significant heterogeneity (p<0.05) between studies for at least one of the localisation categories (correctly localised, not localised, partially localised and incorrectly localised) for all imaging techniques. Possible explanations for this heterogeneity include differing study designs, index test characteristics, reference standards and population characteristics. Test performance was more promising in studies restricted to patients with temporal lobe epilepsy. Ictal single photon emission computed tomography (SPECT) generally had more correctly localising (70--100%) and fewer non-localising (0--7%) scans than other techniques evaluated in patients with temporal lobe epilepsy. Results for computed tomography and interictal SPECT suggest that these tests are relatively poor at localising the seizure focus. Volumeric magnetic resonance imaging (MRI) and position emission tomography (PET) appear promising, and subtraction ictal single photon emission computed tomography co-registered to magnetic resonance imaging (SISCOM) and magnetic resonance spectrosopy (MRS) less promising than ictal SPECT, but these technologies have been assessed in fewer studies. T2 relaxometry was reported in only one small study with inconclusive results. Seventeen studies (33 evaluations) provided sufficient data on the association of a localised scan with outcome following surgery to calculate a relative risk. The majority of evaluations (24/33) suggested that patients with a correctly or partially localised scan had a better outcome following surgery than those with an incorrectly localised or non-localised scan. However, this association was statistically significant in only three studies, two evaluating routine MRI [(relative risk (RR) 2.74, 95% confidence interval (CI): 1.32 to 5.67; RR 1.28, 95% CI: 1.00 to 1.63] and the other SISCOM (RR 2.12, 95% CI: 1.01, 4.44). Nine studies used multivariate analysis to investigate the association of MRI (7 studies), MRS and volumetric MRI (1 study), PET (3 studies), SPECT (1 study) and SISCOM (3 studies) with the outcome following surgery. There was a trend for localisation of abnormalities to be associated with a beneficial outcome. CONCLUSIONS: Due to the limitations of the included studies, the results of this review do little to inform clinical practice, with insufficient evidence regarding effectiveness and cost-effectiveness of imaging techniques in the work-up for epilepsy surgery. Given the inadequacy of existing data, there is a pressing need for studies investigating the utility of imaging techniques in the work up for epilepsy surgery. The most reliable method to achieve this is the RCT, which could examine the single tests or combinations of tests on patient outcome. The authors suggest that it is important that clinicians, patient groups, policy makers and healthcare/research funders meet and debate the most appropriate way to investigate these technologies.