Volume 26, Issue 3 pp. 303-308
Original Research

Diagnostic Performance of Brain MRI in Immune Reconstitution Inflammatory Syndrome

Jared Narvid

Corresponding Author

Jared Narvid

Division of Neuroradiology, Department of Radiology and Biomedical Imaging, University of California, San Francisco, CA

Correspondence: Address correspondence to Jared Narvid, MD, Division of Neuroradiology, 505 Parnassus Avenue, L-352, San Francisco, CA 94143-0628. E-mail: [email protected].Search for more papers by this author
Bhavya Rehani

Bhavya Rehani

Division of Neuroradiology, Department of Radiology and Biomedical Imaging, University of California, San Francisco, CA

Search for more papers by this author
Jason F. Talbott

Jason F. Talbott

Division of Neuroradiology, Department of Radiology and Biomedical Imaging, University of California, San Francisco, CA

Search for more papers by this author
First published: 11 September 2015
Citations: 11

ABSTRACT

BACKGROUND AND PURPOSE

Central nervous system immune reconstitution inflammatory syndrome (CNS-IRIS) significantly negatively impacts the human immunodeficiency virus (HIV)-infected population on combination antiretroviral therapy (cART). We sought to determine the diagnostic performance of several magnetic resonance imaging (MRI) features for CNS-IRIS in a cohort of HIV+ patients recently started on cART.

METHODS

Our radiologic database was searched from January 2003 to September 2014 retrospectively for patients diagnosed with HIV and worsening symptoms on cART. A total of 20 patients with HIV were identified; patients were classified as having CNS-IRIS on the basis of established clinical criteria (8 patients; 12 age- and sex-matched controls). Brain MR images were obtained at a single post-cART timepoint during hospitalization for acute neurologic deterioration and blindly interpreted by two experienced neuroradiologists for the presence of four variables: intrinsic T1 hyperintensity, marginal reduced diffusion, and marginal enhancement or perivascular enhancement.

RESULTS

Although each individual finding showed moderate predictive accuracy, the combination of MR findings demonstrated good test characteristics: sensitivity 88% (confidence interval [CI] 62-98), specificity 79% (58-93), positive predictive value 71% (44-90%), and negative predictive value 83% (CI 52-98%). In addition, this final diagnosis demonstrated good predictive accuracy, area under curve .78 (CI .63-.91), and moderate inter-reader agreement, κ = .55.

CONCLUSIONS

Our findings suggest that although each individual MR finding shows only moderate diagnostic performance, the combined assessment of experienced neuroradiologists has good predictive accuracy. The absence of any described MRI findings makes the diagnosis of CNS-IRIS highly unlikely.

The full text of this article hosted at iucr.org is unavailable due to technical difficulties.