Diagnostic and prognostic biomarkers in immune checkpoint inhibitor-related encephalitis: a retrospective cohort study

Abstract

Background: Immune checkpoint inhibitor-related encephalitis (ICI-encephalitis) is not well characterised and diagnostic and prognostic biomarkers are lacking. We aimed to comprehensively characterise ICI-encephalitis and identify diagnostic biomarkers and outcome predictors.

Methods: This retrospective observational study included all patients with ICI-encephalitis studied in the French Reference Centre on Paraneoplastic Neurological Syndromes (PNS) and Autoimmune Encephalitis (2015-2023). ICI encephalitis was considered definite in case of inflammatory findings at paraclinical tests and/or well-characterised neural antibodies. Predictors of immune-related adverse event (irAE) treatment response, defined as a Common Terminology Criteria for Adverse Events v5.0 grade < 3 at any time after therapeutic intervention, were assessed by logistic regression analysis, and predictors of mortality by Cox regression analysis. Neurofilament light chain (NfL) was measured by enzyme-linked immunosorbent assay.

Findings: Sixty-seven patients with definite encephalitis were identified (median age, 69 years; 66% male). A focal syndrome was observed in 43/67 patients (64%; limbic encephalitis, cerebellar ataxia, and/or brainstem encephalitis), while 24/67 (36%) had meningoencephalitis, a non-focal syndrome with altered mental status (22/24 patients, 92%) and pleocytosis (24/24 patients, 100%). Patients with focal encephalitis more frequently had abnormal brain MRI (26/42, 62% versus 8/24, 33%, p = 0.025), PNS-related antibodies (36/43, 84% versus 1/24, 4%, p < 0.001), and neuroendocrine cancers (22/43, 51% versus 1/24, 4%; p < 0.001) than patients with meningoencephalitis. Focal encephalitis patients had a lower rate of irAE treatment response (7/39, 18%) and higher mortality (27/43, 63%) compared to meningoencephalitis patients (12/22, 77% and 5/24, 21%, respectively, p < 0.001 each). PNS-related antibodies were associated with less irAE treatment response, independently of age, sex, and baseline severity (adjusted OR 0.05; 95%CI [0.01; 0.19]; p < 0.001) as well as higher mortality, independently of age and cancer type (adjusted HR 5.07; 95% CI [2.12; 12.12]; p < 0.001). Serum NfL discriminated patients with definite ICI-encephalitis (n = 27) from cancer-matched controls (n = 16; optimal cut-off >273.5 pg/mL, sensitivity 81%, specificity 88%, AUC 0.87, 95% CI [0.76; 0.98]) and irAE treatment responders (n = 10) from non-responders (n = 17, optimal cut-off >645 pg/mL, sensitivity 90%, specificity 65%; AUC 0.75, 95% CI [0.55; 0.94]).

Interpretation: ICI-encephalitis corresponds to a set of clinically-recognisable syndromes. Patients with focal encephalitis, PNS-related antibodies, and/or higher serum NfL have low irAE treatment response rates. Research is needed on the underlying immunopathogenesis to foster therapeutic innovations.

Funding: Agence Nationale de la Recherche.

Keywords: Adverse events; Autoimmune encephalitis; Biomarkers; Cancer; Cancer immunotherapy; Encephalitis; Immune checkpoint inhibitors; Neural antibodies; Neurological immune-related adverse events; Paraneoplastic neurological syndromes; Prognostic factors.

Fig. 1

Representative brain MRI findings. Bilateral, left predominant temporo-insular T2/FLAIR hyperintensity (a and b) and temporo-mesial contrast enhancement (c) in a 44-year-old man with antibody-negative limbic encephalitis. Bilateral temporo-mesial T2/FLAIR hyperintensity in a 63-year-old woman with anti-Hu limbic encephalitis (d). Moderate cerebellar atrophy and T2/FLAIR superior cerebellar peduncles hyperintensity in a 76-year-old man with anti-Dach1 cerebellitis (e). Leptomeningeal (perivascular spaces) (f) and epdendymal (g) contrast enhancement, and T2/FLAIR left optic nerve hyperintensity (arrow, h) in a 70-year-old patient with anti-GFAP meningoencephaltis.

Fig. 2

Syndrome-antibody correlations in patients with definite immune checkpoint inhibitor-encephalitis. Syndromes observed in each antibody-defined group are represented at the individual level. The intersections between syndromes and antibodies in each patient are represented by dots, connected by a line in case of overlapping syndromes. Anti-MOG antibodies were negative in all 16 patients tested (12 meningoencephalitis, 4 focal encephalitis).

Fig. 3

Outcome and prognostic factors. CTCAE severity at baseline (M0) and at 3 (M3), 6 (M6), 9 (M9), and 12 months (M12) in all patients with definite encephalitis (n = 67), and in the subgroups of patients with focal encephalitis (n = 43) and with meningoencephalitis (n = 24). CTCAE v 5.0: grade 1 = asymptomatic or mild; grade 2 = minimal, non-invasive intervention indicated; grade 3 = severe, not immediately life-threatening; grade 4 = life-threatening, urgent intervention indicated; grade 5 = death. Abbreviations: CTCAE, Common Terminology Criteria for Adverse Events; M, month.

Fig. 4

Light chain neurofilaments: diagnostic performance for definite immune-checkpoint inhibitor-encephalitis and comparison with Herpes Simplex Virus encephalitis, anti-LGI1 encephalitis, anti-Hu encephalitis, and cancer-matched controls. Comparison of serum NfL in patients with definite ICI-encephalitis, Herpes Simplex Virus encephalitis, anti-LGI1 encephalitis, anti-Hu encephalitis and cancer-matched controls (a); ∗∗∗∗p ≤ 0.0001; ∗p ≤ 0.05; ns, p > 0.05. Receiver operating characteristic (ROC) curves of serum NfL levels to discriminate patients with definite ICI encephalitis from cancer-matched controls (b). Abbreviations: AUC, area under the ROC curve; CI, 95% confidence interval; HSV-E, Herpes Simplex Virus encephalitis; Hu-E, anti-Hu encephalitis; ICI-E, ICI-encephalitis; LGI1-E, anti-LGI1 encephalitis.

Fig. 5

Lightchain neurofilaments: comparison between irAE treatment responders and non-responders. Comparison of NfL serum levels between irAE treatment responders and non-responders (a), and receiver operating characteristic (ROC) curves of serum NfL levels to discriminate irAE treatment responders and non-responders (b); irAE treatment responders (n = 10) had lower serum NfL levels (median 392.5 pg/mL, range 32–1606) compared to non-responders (n = 17, median 689 pg/mL, range 70–13522, p = 0.035). Comparison of NfL CSF levels between irAE treatment responders and non-responders (c), and receiver operating characteristic (ROC) curves of CSF NfL levels to discriminate treatment responders and non-responders (d); NfL levels were lower in the CSF of irAE treatment responders (n = 8, median 8400 pg/mL, range 1186–359,556) compared with non-responders (n = 8, median 34,200 pg/mL, range 4669–241,762), but this difference was not significant (p = 0.083). Abbreviations: AUC, area under the ROC curve; CI, 95% confidence interval; irAEs, immune-related adverse events.