Functional MRI and electrical stimulation mapping for language localization: A comparative meta-analysis.

OBJECTIVE

This systematic review and meta-analysis evaluated the diagnostic validity of functional magnetic resonance imaging (fMRI) compared to electrical stimulation mapping (ESM) for pre-surgical language mapping.

METHODS

A structured literature search was performed and studies with electrode-level data comparing fMRI and ESM for language localization were analyzed. Outcome measures included pooled estimates of diagnostic odds ratio (DOR), sensitivity, and specificity. Sources of heterogeneity were explored with a meta-regression framework.

RESULTS

Nine studies were included in the analysis having 5-40 patients with mean age 11.3-43.4 years. Verb generation and picture naming were the most common fMRI tasks, while picture naming was the most common ESM task. Sensitivity (0.37-0.95), specificity (0.36-0.97), and DOR (1.9-44.6) for fMRI compared to ESM varied widely across studies with statistically significant heterogeneities. Pooled estimates were: sensitivity 0.71 (95% confidence interval 0.54, 0.83), specificity 0.74 (0.58, 0.85), and DOR 7.0 (3.5, 13.8), from a valid meta-analysis (area under the summary receiver-operating-curve 0.78). fMRI was noted to have higher sensitivity in studies using higher maximal ESM currents. fMRI variables could not be included in the meta-regression because of substantial methodological differences among studies.

CONCLUSIONS

fMRI is moderately sensitive and moderately specific for language localization compared to ESM in well-designed studies. However, because the confidence limits for sensitivity and DOR are close to the line of no effect, and there is high unmeasured heterogeneity, fMRI cannot perhaps be used as the only modality for language localization to inform neurosurgical decisions at present.