Commentary
Magnetoencephalography (MEG) for the localization of the epileptogenic zone entered clinical practice in the late 1990s and was routinely used by the mid-2000s.1 Since that time, the adoption of this technology by the epilepsy community has been steady but slow. Although there are currently over 25 clinical MEG centers in the United States performing epilepsy studies, many academic centers and large epilepsy practices do not have on-site MEG access in the same manner as say, the now ubiquitous 3T MRI scanner which became commercially available in 1998.2 The fact that the MEG scanner needs to be housed in a special magnetically shielded room and the entire system costs over US $2M (personal communication) may factor into the slow adoption rate. However, the statement that MEG provides added value to the presurgical workup of select patients with epilepsy should not be a controversial one and not likely to be the reason for the slow adoption. From a scientific perspective, there have been numerous prospective studies3,4 and retrospective large series5 which have established the efficacy and safety of MEG in localizing the epileptogenic zone.
The work of Rampp et al6 adds to the existing literature by presenting the largest case series of MEG studied for the localization of epileptogenic foci. This is a single-center retrospective observational cohort study spanning almost 30 years and reviewing 1274 MEG tests. The hypothesis was that MEG supports patient selection, contributes to the identification of the epileptogenic zone in presurgical evaluation, and impacts long-term seizure outcomes after epilepsy surgery. All recordings, source imaging, and interpretation were performed prospectively before surgery. Over the span of the study, 3 different MEG systems were used: a 37-channel radiometer system; a 2 × 37 channel axial gradiometer system or a 148-channel axial gradiometer system; and a 248-channel whole-head magnetometer. Multiple source localization methods with multiple volume conductor models (single sphere, multiple spheres, and boundary element method) and multiple inverse solutions algorithms (LORETTA/sLORETTA, minimum norms, and CLARA) were used with Curry (Compumedics; Charlotte, NC) or BESA (BESA GmbH; Grafelfing, Germany) software. Association of seizure outcome after surgery with MEG/focus hypothesis concordance, as well as MEG resection, was evaluated. Magnetoencephalography detected pathologic spikes/sharp waves in 72% of the recordings, which is similar to other observational studies. Magnetoencephalography findings were “concordant” with the presurgical consensus (derived from video-electroencephalogram, structural and functional testing) in 50% of the patients, “consistent” in 32% and “discordant” in 18%. The percentage was higher in temporal lobe epilepsy cases compared to extra-temporal lobe epilepsy cases, which is not surprising and in line with what is seen in clinical practice. Approximately 40% of the cohort went on to epilepsy surgery. Follow-up data were available at the 1- and 5-year mark for many patients, with a few even up to 20 years. Complete resections of focal MEG localizations were significantly related to Engel 1 outcome. Complete resection of an epileptogenic lesion was also related to a better outcome, compared to cases with partial or no resection. Interestingly, the association of Engel 1 outcomes with complete MEG resections was stronger in non-lesional versus lesional cases.
The main strength of this paper is the sheer volume of the clinical material—almost 1300 MEG studies in 1000 consecutive patients evaluated as part of the workup for possible epilepsy surgery. A recent meta-analysis7 found 23 papers that associated MEG spiking to surgical resection site and then to surgical outcome. The total N was 440 in those 23 studies. When one notes that 405 patients underwent epilepsy surgery in Rampp et al, the contribution to the literature is immediately obvious. Adding to this, the longitudinal duration of this study spanning 28 years and a trial conducted at one of the oldest and most established MEG centers in Europe, and you have an impressive collection of carefully acquired data. If there were any doubts regarding the amount of available scientific data on MEG as a diagnostic tool for epilepsy, this study should dispel those concerns.
And yet, some of the strengths are also reflected in the limitations. During the long duration of the study, technological advancements in MEG hardware and software at the study site changed the study methodology. Over the course of study, 3 different MEG sensors acquired the data, ranging from a 37-channel partial hemisphere system to a 248-channel whole-head system. Multiple localization methods were utilized. While the authors did perform subgroup analyses and report that the results were not significantly different from era to era, one cannot help but wonder whether this unexpected finding is generalizable to all MEG centers or more due to the skill and expertise of the staff at Erlangen. The structure of this study also is designed to highlight the utility of MEG as a technology rather than as a specific method, as multiple methods were used. Some of the data presented in this paper have been previously described.8 In a way, the paper almost acts as a review of the Erlangen experience over 3 separate eras.
The findings in this paper are in line with other studies on the subject and do not break significant new ground. A single-center prospective study of MEG (3) found a positive predictive value of 78% and a corrected negative predictive value of 64% with respect to the surgical outcome (compared to 83% and 65%, respectively, for the current study). A retrospective cohort study of 132 patients5 with a similar design to the current study was able to record interictal discharges in 78% of patients (compared to 72%) and found localizing MEG spikes predicted seizure freedom with a diagnostic odds ratio of 5.1 (compared to 8.4).
Rampp et al have contributed to firmly establishing MEG as one of the diagnostic tools in the presurgical evaluation of drug-resistant epilepsy. The next step for the field is to elevate the technology from “it would be nice to have in my backyard” to “I need this at my center”. What is needed to make this jump are new findings that would make MEG essential in the diagnostic workup for epilepsy surgery. One potential major advancement is for MEG to identify noninvasive biomarkers of the epileptogenic focus with high positive and negative predictive values. Coincidentally, in the very next issue of Brain after Rampp et al is another MEG paper9 reporting to be the first noninvasive study to detect high-frequency oscillations, address the efficacy of high-frequency oscillations over the different neural oscillatory frequencies, localize them and clinically validate them with the postsurgical outcome in patients with medically refractory epilepsy. Further studies that replicate and validate the finding that high-frequency oscillations detected by MEG significantly improve the presurgical assessment and can predict the postsurgical outcome would make MEG an indispensable technology in every epilepsy center.  By Jerry J. Shih