AbstractBackgroundBiomarkers are lacking that can discriminate frontotemporal lobar degeneration (FTLD) with tau (FTLD‐tau) from TDP‐43 (FTLD‐TDP). In FTLD patients with autopsy‐ or mutation‐confirmed pathology, we tested the glial fibrillary acidic protein (GFAP) to neurofilament light chain (NfL) ratio (GFAP/NfL) in plasma. We validated application of GFAP/NfL ratio in an independent sample of frontotemporal dementia (FTD).MethodsTraining sample was composed of 105 FTLD‐TDP and 44 FTLD‐tau participants, with pathology confirmed by either autopsy or pathogenic mutations (MAPT, C9orf72, GRN, TARDBP). Average accumulations of postmortem tau, TDP‐43, and amyloid across neocortical, limbic and brainstem regions were scored on a 0‐3 scale. Regression tested differences in analytes (log transformed) across FTLD‐tau and FTLD‐TDP, covarying for age, interval to death, and sex. Regression tested association of analytes (log transformed) with postmortem tau and TDP‐43 accumulation, covarying for postmortem amyloid‐β and plasma‐to‐death interval. Receiver operating characteristic (ROC) analyses and area under the curve (AUC) analyses were performed in patients with negligible secondary pathology. We defined an optimal cut‐point to discriminate FTLD‐tau from FTLD‐TDP using plasma GFAP/NfL. We tested the defined cut‐point in an independent validation sample of FTD patients with a clinical diagnosis of progressive supranuclear palsy (PSP) with likely tau (FTD‐tau; n=33) or diagnosis of amytrophic lateral sclerosis (ALS) with likely TDP‐43 (FTD‐TDP; n=35).ResultsPlasma GFAP was lower in FTLD‐TDP than FTLD‐tau (β=‐0.24, p=0.021) with small effect size (partial η2=0.082). Plasma NfL was higher in FTLD‐TDP than FTLD‐tau (β=0.57, p=1.0e‐05) with medium effect size (partial η2=0.24). GFAP/NfL was significantly lower in FTLD‐TDP than FTLD‐tau (β=‐0.82, SE=0.13, p=2.2e‐09) with large effect size (partial η2=0.39). Higher plasma GFAP was associated with higher postmortem tau accumulation (β=0.15, p=0.032). Higher levels of plasma NfL were associated with increased TDP‐43 burden (β=0.2, p=0.024). ROC revealed excellent discrimination by plasma GFAP/NfL in the autopsy/test sample (AUC=0.89). In the validation sample, plasma GFAP/NfL had good overall discrimination accuracy (AUC=0.85); there was 0.79 sensitivity and 0.77 specificity when applying the training‐derived GFAP/NfL threshold.ConclusionPlasma GFAP and NfL relate to pathological accumulation of tau and TDP‐43, respectively. The plasma ratio of GFAP/NfL is a promising candidate to discriminate FTLD‐tau from FTLD‐TDP.