AAN Guideline: Use of fMRI in the Presurgical Evaluation of Patients with Epilepsy (2017)

This topic summarizes the 2017 AAN practice guideline by Szaflarski et al. on the use of functional MRI (fMRI) in the presurgical evaluation of patients with epilepsy, addressing language and memory lateralization, outcome prediction, and potential replacement of the intracarotid amobarbital procedure (Wada test).

Guideline Overview

Source & Scope

• Organization: AAN Guideline Development, Dissemination, and Implementation Subcommittee
• Published: Neurology 2017;88:395–402 (January 24, 2017)
• Lead authors: Szaflarski JP, Gloss D, Binder JR, Gaillard WD, Golby AJ, Holland SK, Ojemann J, Spencer DC, Swanson SJ, French JA, Theodore WH
• Endorsed by: American College of Radiology (September 2016) and American Epilepsy Society (December 2016)
• Evidence base: 37 articles selected from 172 reviewed; ≥2 panelists reviewed each article; studies with n < 15, case reports, meta-analyses, and editorials excluded
• Evidence rating: 2004 AAN diagnostic and prognostic classification scheme (Class I–III)

Six Clinical Questions Addressed

1. Is fMRI comparable with IAP for measuring language lateralization?
2. Can fMRI predict postsurgical language outcomes in patients with epilepsy undergoing brain surgery?
3. Is fMRI comparable with IAP for measuring memory lateralization?
4. Can fMRI predict postsurgical verbal memory outcomes in patients undergoing temporal lobectomy?
5. Can fMRI predict postsurgical nonverbal (visuospatial) memory outcomes in patients undergoing medial temporal lobectomy?
6. Is there sufficient evidence for fMRI to replace the IAP (Wada test) in presurgical evaluation?

Key Abbreviations

Q1: fMRI vs IAP for Language Lateralization

Evidence Summary

• Class I & II studies available — Class III studies not discussed for this question
• Individual patient data meta-analysis performed on Class I and II studies for MTLE
• Additional data: 6 Class II studies + 12 Class III studies that did not specify medial vs lateral temporal localization

Concordance by Epilepsy Type

Temporal Lobe Epilepsy (TLE)

• Class I study (Janecek et al., 229 patients): fMRI concordant with IAP in 81/91 (89%) for right and 82/97 (85%) for left medial temporal foci
• Presence of medial temporal sclerosis did not affect concordance rate
• Class II study (1): Concordance 12/14 (86%) in patients with MTLE and 3/3 (100%) in medial temporal tumors
• Class II study (2): Reading task only — concordance 26/31 for all patients and 9/13 for medial temporal seizure onset
• Class II study (3): Excellent concordance — 17/17 for MTLEs and 4/4 for medial temporal tumors

Extratemporal Epilepsy

• Class I study (Janecek et al.): Concordance in 34/41 (83%) of extratemporal cases
• Class II study (1): 40 patients with extratemporal epilepsies — concordance 8/11 for all IAP results, 8/10 if bilateral language excluded
• Class II study (2): 26 patients with epilepsy — 100% concordance between fMRI and IAP in 5/5 extratemporal lobe epilepsy patients

Mixed / Unspecified Foci

• 1 Class II study in 20 patients with epilepsy or brain tumors — 86% overall correlation with IAP (sentence task and synonym task)
• 1 Class II study in 51 patients — moderate correlation (r = 0.68; p < 0.0001) between IAP and fMRI
• 1 Class II study in 38 patients — concordance in 1/2 extratemporal cases

Meta-Analysis Concordance Rates (Definite Right or Left IAP Results Only)

Class I & II fMRI–IAP Language Lateralization Comparison Studies

^a Combined for all tasks.

Conclusions & Recommendation

• fMRI possibly provides language lateralization information concordant with IAP in 87% of medial temporal cases and 81% of extratemporal cases
• Insufficient data for temporal tumors or lateral temporal cases

Q2: fMRI for Predicting Postsurgical Language Outcomes

Evidence Summary

• 1 Class II study (44 patients with left/right TLE + hippocampal sclerosis) and 1 Class III study (56 patients with left/right MTLE)

Class II Study Findings (Bonelli et al., 2012)

• Patients with left TLE — strong left frontal activation predicted greater postresection language decline:
  • Sensitivity: 100%
  • Specificity: 33%
  • Positive predictive value: 60%
• Postresection performance depended on greater right frontal language activation shift

Class III Study Findings (Sabsevitz et al., 2003)

• Stronger leftward lateralization in temporal lobe ROI during semantic decision task → greater postoperative naming decline
• Sensitivity: 100%; Specificity: 73%; PPV: 81%
• Temporal lobe LI correlation to postoperative Boston Naming Test: r = −0.64; p < 0.001
• Same study compared IAP prediction accuracy: Sensitivity 92%, Specificity 45%, PPV 67% — lower than fMRI

fMRI vs IAP for Predicting Language Outcome

Conclusion & Recommendation

• fMRI is possibly effective in predicting postsurgical language deficits (1 Class II + 1 Class III study)
• Evidence for IAP language outcome prediction is also limited (1 Class II + 1 Class III study)

Q3: fMRI vs IAP for Memory Lateralization

Evidence Summary

• 2 Class II studies and 2 Class III studies available
• Results are mixed — one study supports concordance, one does not

Class II Study 1 (Dupont et al., 2010)

• 67 patients with TLE
• fMRI paradigm: contrast between novel visual scenes and meaningless visual patterns
• Significant correlation between hippocampal fMRI LI and IAP memory LI: r = 0.31; p = 0.007

Class II Study 2 (Binder et al., 2010)

• 25 patients with TLE
• No significant correlation between fMRI asymmetry measure and medial temporal IAP memory LI: r = 0.152; p = 0.47

Class III Studies

• Study 1 (18 patients with TLE): fMRI paradigm contrasting novel vs studied pictures — MTL region LI correlated with IAP memory LI (r = 0.49; p = 0.049)
• Study 2 (30 patients with left TLE): Number of activated left MTL voxels positively correlated with left IAP memory score (Spearman r = 0.60; p < 0.01)

Conclusion & Recommendation

• In patients with MTLE, Class II evidence suggests fMRI is comparable with IAP for memory lateralization
• The conflicting data from one study may be related to a relatively high dose of sodium amobarbital used in the IAP

Q4: fMRI for Predicting Postsurgical Verbal Memory Outcomes

Evidence Summary

• 9 Class II studies and 3 Class III studies addressed this question
• This is the most robust evidence base in the guideline and produced the only Level B recommendation

Key Class II Study Findings

• Study 1 (Binder et al., 2008): 122 patients with TLE (60 left) — 50% of variance in postsurgical verbal memory outcome in left TLE was explained by preoperative neuropsychological testing; fMRI explained an additional 10% of variance (p ≤ 0.003)
• Study 2 (Binder et al., 2010): Lateralization of hippocampal activation during picture encoding was NOT predictive of postresection verbal memory outcomes (in contrast to language network lateralization)
• Study 3 (Bonelli et al., 2010): 54 patients with TLE (29 left) using word-encoding fMRI paradigm — degree of fMRI asymmetry toward left correlated with postsurgical verbal memory decline (p = 0.028)
• Study 4 (Bonelli et al., 2010): Model including left fMRI activation during delayed recognition, side of seizure onset, and preoperative verbal memory score → correctly predicted worsening of verbal memory in 90% of patients
• Study 5 (Sidhu et al., 2015): 50 patients with TLE (23 left) — increasing left lateralization in frontotemporal verbal memory network → associated with postresection verbal memory decline in left TLE (r = 0.44; p = 0.037); NOT observed in right TLE
• Study 6: 21 patients with left or right TLE — greater functional connectivity between hippocampus and Brodmann area 22 (superior temporal gyrus) before resection → associated with decreased verbal memory after surgery

Conclusion & Recommendation

• fMRI leftward activation asymmetry during encoding of verbal material (whether measured in the MTL or language network) probably predicts verbal memory decline after left MTL surgery
• Evidence comes from 9 Class II studies using different methods — converging findings strengthen the conclusion

Q5: fMRI for Predicting Postsurgical Nonverbal (Visuospatial) Memory Outcomes

Evidence Summary

• Only 1 Class II study addressed this question

Class II Study (Bonelli et al., 2010)

• 72 patients (68 with unilateral hippocampal sclerosis)
• Asymmetry of face recognition–related activation was the best predictor of visual-spatial decline after surgery
• Greater right anterior MTLE activation for encoding faces → correlated with greater visual memory decline after right ATL resection
• Correlation: r = 0.47; p = 0.02

Conclusion & Recommendation

• fMRI activation asymmetry during nonverbal (scene and face recognition) memory tasks is possibly predictive of nonverbal memory decline after MTL surgery (1 Class II study)

Q6: Can fMRI Replace the IAP (Wada Test)?

Advantages of fMRI over IAP

• Lower risk — noninvasive (no arterial catheterization)
• Lower cost
• Greater potential for localization of function (spatial resolution)
• Repeatable — can be performed multiple times without risk

Language Lateralization: fMRI as Replacement

• Several Class I–III studies support fMRI for language mapping
• Concordance is generally high but not 100% (see table above: 87% medial temporal, 81% extratemporal)
• Evidence for fMRI to predict language outcome (not just lateralization) is limited to 1 Class II + 1 Class III study
• Data on the ability of IAP to predict language outcome are also limited

Conclusion

• fMRI is possibly an effective method of lateralizing language functions and may be a suitable replacement for IAP for this purpose
• Data on ability of fMRI to predict language outcomes are limited

Memory: fMRI as Replacement for IAP

• Concordance between IAP memory asymmetry and fMRI MTL activation asymmetry is modest
• The IAP itself has a limited ability to predict postoperative verbal memory change
• Key finding: In 60 patients who had left ATL surgery, fMRI language LI was more strongly correlated with verbal memory change (r = 0.44; p < 0.001) than IAP memory asymmetry (r = 0.30; p < 0.05)
• A multivariate prediction model including preoperative memory score, age at seizure onset, and fMRI LI was NOT improved by adding IAP asymmetry scores
• 9 Class II studies (including one showing fMRI LI is more accurate than IAP) support that fMRI may be an alternative to IAP for predicting material-specific verbal memory change

Conclusion

• fMRI of language and verbal memory lateralization may be an alternative to IAP memory testing for prediction of verbal memory outcome in MTLE
• fMRI is NOT yet established as a replacement for the IAP for prediction of global amnesia in patients undergoing unilateral ATL resection
• Global amnesia is rare after unilateral temporal lobe surgery and occurs mainly with preexisting contralateral MTL dysfunction

Clinical Context & Limitations

Limitations of the Evidence

• Evidence derived from relatively small patient samples with heterogeneous characteristics
• Some studies underpowered or susceptible to random variation
• Few studies examined fMRI ability to predict language outcomes
• No multicenter studies assessing replicability of fMRI methods across centers
• Vast majority of data from adults with TLE and minimal structural lesions
• No Class I or II studies that solely address fMRI in children/adolescents or comparison with IAP in younger age ranges
• Limited data on patients with extratemporal foci and larger lesions

Methodologic Variability

• Magnetic field strength varied across studies
• Techniques for analysis of raw data varied
• Thresholding methods, ROI selection, and lateralization methods differed
• Effect of these variables on data quality and validity is currently unknown
• The guideline assumes published standards are followed for conducting clinical fMRI studies

Important Clinical Considerations

• fMRI is a complex diagnostic procedure requiring advanced technical expertise in imaging and expert interaction with patients
• Clinicians must select activation tasks appropriate to the patient's ability and clinical aims
• Patient compliance with activation tasks is a prerequisite for valid fMRI results
• Task design, analysis methods, and epilepsy type (temporal vs extratemporal, lesional vs nonlesional) must be considered when interpreting results
• Patients with lesional epilepsy — only small numbers included in prior studies; variable lesion size/location is a concern
• Clinicians should carefully advise patients of the risks and benefits of fMRI vs IAP during discussions concerning choice of modality

Recommendations for Future Research

• Studies comparing fMRI and IAP for language and memory outcome prediction
• Studies comparing fMRI ability to predict outcomes with various surgical treatments
• Studies comparing various fMRI language and memory tasks for lateralization, IAP agreement, and outcome prediction
• Studies comparing various fMRI analysis methods using postsurgical outcomes as standards
• Multicenter studies assessing replicability
• Studies targeting extratemporal and lesional epilepsy
• Studies targeting pediatric epilepsy populations

Summary of All Recommendations by Evidence Level

Quick Reference: Evidence Level Definitions (AAN Scheme)