Paraneoplastic autoantibody panels: sensitivity and specificity, a retrospective cohort

Abstract

Background: Experts in the autoimmune paraneoplastic field recommend autoantibody testing as “panels” to improve the poor sensitivity of individual autoantibodies in detecting paraneoplastic neurological syndromes (PNS). The sensitivity of those panels was not reported to date in a fashion devoid of incorporation bias. We aimed to assess the collective sensitivity and specificity of one of the commonly used panels in detecting PNS. Methods: A single-centered retrospective cohort of all patients tested for paraneoplastic evaluation panel (PAVAL; test ID: 83380) over one year for the suspicion of PNS. Case adjudication was based on newly proposed diagnostic criteria in line with previously published literature, but modified to exclude serological status to avoid incorporation bias. Measures of diagnostic accuracy were subsequently calculated. Cases that failed to show association with malignancy within the follow-up time studied, reflecting a possibly pure autoimmune process was considered paraneoplastic-like syndromes. Results: Out of 321 patients tested, 51 patients tested positive. Thirty-two patients met diagnostic criteria for paraneoplastic/paraneoplastic-like syndromes. The calculated collective sensitivity was 34% (95% CI: 17–53), specificity was 86% (95% CI: 81–90), Youden's index 0.2 and a positive clinical utility index 0.07 suggesting poor utility for case-detection. Conclusion: This is the first reported diagnostic accuracy measures of paraneoplastic panels without incorporation bias. Despite recommended panel testing to improve detection of PNS, sensitivity remains low with poor utility for case-detection. The high-calculated specificity suggests a possible role in confirming the condition in difficult cases suspicious for PNS, when enough supportive evidence is lacking on ancillary testing.

KEYWORDS:

• sensitivity and specificity
• paraneoplastic autoantibodies
• neurological paraneoplastic syndromes

Acknowledgements

We thank Aaron Bubolz for helping with initial data collection, Firas Al Badarin for his insightful suggestions and manuscript review, Marilyn Rymer, Lowell Tilzer, Erica Howe and Chris Wittkopp for their help obtaining the database.

Declaration of Interest

Drs Albadareen, Gronseth, Goeden, Sharrock, Lechtenberg and Wang have no financial disclosures related to this work.

This study was not supported with any funding sources.

Supplementary material

Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/00207454.2016.1207644