Evaluating finger-prick blood collection for remote quantification of neurofilament light in neurological diseases

Abstract

Promising blood-based biomarkers of neuropathology have emerged with potential for therapeutic development and disease monitoring. However, these tools will require specialist tertiary services for integration into clinical management. Remote sampling for biomarker assessment could reduce burden of in-person clinical visits for such tests as well as increasing the sampling frequency and patient geographical outreach. Here, we evaluated a finger-prick blood collection approach for remote quantification of neurofilament light (NfL), a candidate blood-based biomarker evident in various neurological disorders, and other exploratory markers of neuronal injury and neuroinflammation (GFAP, tau). Matched samples from venepuncture and finger-prick were collected and processed into plasma and/or serum to directly compare analyte levels from a multi-disease discovery cohort (n = 54 healthy controls; n = 57 Huntington's disease (HD); n = 34 multiple sclerosis; n = 7 amyotrophic lateral sclerosis; n = 11 Parkinson's disease), and a HD confirmatory cohort (n = 57 healthy controls; n = 64 HD). Two delayed processing conditions were compared, three- and seven-day delay, simulating ambient shipment. Capillary NfL and GFAP concentrations were equivalent to those in venous serum and plasma in the multi-disease discovery cohort and HD confirmatory cohort. Only NfL remained stable after a seven-day processing delay in both venous and capillary serum samples. Using NfL concentrations from capillary blood, we replicated previously published disease group differences measured in venous blood. This data supports our finger-prick approach for remote collection and quantification of NfL. With the widespread applications for NfL across the spectrum of neurological disorders, this has the potential to transform disease monitoring, prognosis, and therapeutic development within clinical practice and research.

Keywords: Biomarkers; Blood; Finger-prick; Neurodegenerative disorders; Neurofilament light; Remote sampling.

Fig. 1

Study design and experimental objectives. A Schematic showing capillary sample collection method: 1. Finger-prick performed using a fast-flow lancet, 2. Blood milked from finger, 3. Up to 600 µL whole blood collected into two LiHep and/or SST microtainers, 4. ~ 300 µL capillary plasma/serum generated after processing per microtainer. B Experimental objective A.0.0: matched capillary (microtainer) and venous (vacutainer) samples were collected in serum and plasma tubes and processed on the day of collection to compare analytes between collection methods (finger-prick/venepuncture) and sample types (plasma/serum) in both the multi-disease discovery cohort and HD confirmatory cohort. C Experimental objective B.1.a: assessing the impact of 3-day processing delay in plasma. D Experimental objective B.2.a: assessing the impact of 3-day processing delay in serum. E Experimental objective B.1.b: assessing the impact of 7-day processing delay in plasma. F Experimental objective B.2.b: assessing the impact of 7-day processing delay in serum. For delayed processing experiments, C–F two tubes were collected from both finger-prick and venepuncture for either plasma or serum and one tube of each collection type was processed on day zero and the other on the respective processing delay day after shaking at room temperature. Green tubes indicate plasma and yellow tubes indicate serum. ALS amyotrophic lateral sclerosis; HD Huntington’s disease; MS multiple sclerosis; PD Parkinson’s disease; Pre-HD premanifest HD; RT room temperature

Fig. 2

A.0.0: Analyte concentrations in venous versus capillary and plasma versus serum. Analyte concentrations across different collection methods and sample types from experiment A.0.0 for NfL (A–E), GFAP (F–J), and tTau (K–O) in healthy controls, pre-HD, manifest-HD, PD, RRMS, PPMS, SPMS, ALS-slow, ALS-fast. Blue lines represent the line of equality (y = x). Grey lines represent the linear regression fit of the data with 95% confidence interval shaded regions. R² and p values were generated from regression models comparing the two collection types in each panel. The Bonferroni threshold for this experiment was 0.0025 (20 comparisons) and all statistics which reached significance below this are highlighted in bold. NfL, GFAP, and tTau concentrations were natural log-transformed. ALS amyotrophic lateral sclerosis; GFAP glial fibrillary acidic protein; Haem haemoglobin; HD Huntington’s disease; NfL neurofilament light; PD Parkinson’s disease; PPMS primary progressive multiple sclerosis; RRMS relapsing–remitting multiple sclerosis; SPMS secondary progressive multiple sclerosis; tTau total tau

Fig. 3

B.2.a: Analyte concentrations in venous/capillary serum with and without 3-day delay in processing. Analyte concentrations in serum after zero-day and three-day delay in processing from experiment B.2.a. in the multi-disease discovery cohort for total protein (A–E), haemoglobin (F–J), NfL (K–O), GFAP (P–T), and tTau (U–Y) from venous and capillary serum in healthy controls, pre-HD, manifest-HD, PPMS, SPMS. Blue lines represent the line of equality (y = x). Grey lines represent the linear regression fit of the data with 95% confidence interval shaded regions. R² and p values were generated from regression models comparing the impact of delayed processing in each panel. The Bonferroni threshold for this experiment was 0.005 (10 comparisons) and all statistics which reached significance below this are highlighted in bold. NfL and tTau concentrations were natural log-transformed. GFAP glial fibrillary acidic protein; Haem haemoglobin; HD Huntington’s disease; NfL neurofilament light; PPMS primary progressive multiple sclerosis; SPMS secondary progressive multiple sclerosis; tTau total tau

Fig. 4

B.2.b: Analyte concentrations in venous/capillary serum with and without 7-day delay in processing. Analyte concentrations in serum after zero-day and seven-day delay in processing from experiment B.2.b for total protein (A–E), haemoglobin (F–J), NfL (K–O), GFAP (P–T), and tTau (U–Y) from venous and capillary serum in healthy controls, pre-HD, manifest-HD, and SPMS. Blue lines represent the line of equality (y = x). Grey lines represent the linear regression fit of the data, with 95% confidence interval shaded regions. R² and p values were generated from regression models comparing the impact of delayed processing in each panel. The Bonferroni threshold for this experiment was 0.005 (10 comparisons) and all statistics which reached significance below this are highlighted in bold. NfL and tTau concentrations were natural log-transformed. GFAP glial fibrillary acidic protein; Haem haemoglobin; HD Huntington’s disease; NfL neurofilament light; SPMS secondary progressive multiple sclerosis; tTau total tau

Fig. 5

NfL group comparisons for all day-zero samples in the multi-disease discovery cohort. Disease group comparison of NfL concentrations in the multi-disease discovery cohort from A venous plasma, B venous serum, C capillary plasma, and D capillary serum from all baseline data for healthy controls, PPMS, RRMS, SPMS, PD, Pre-HD, manifest-HD, ALS-slow and ALS-fast patients. p values were generated from multiple linear regressions. The Bonferroni threshold for this experiment was 0.0025 (20 comparisons), and all statistics which reached significance below this are highlighted in bold. NfL values are natural log-transformed. ALS amyotrophic lateral sclerosis; HD Huntington’s disease; Man manifest; MS multiple sclerosis; NfL neurofilament light; PD Parkinson’s disease; PP primary progressive; Pre premanifest; RR relapsing–remitting; SP secondary progressive

Fig. 6

Replicating NfL consistency and disease phenotype across sample type in HD confirmatory cohort. NfL analyte concentrations across different collection methods and sample types from the HD confirmatory cohort (A–E) in healthy controls and pre-HD individuals. Blue lines represent the line of equality (y = x). Grey lines represent the linear regression fit of the data, with 95% confidence interval shaded regions. R² and p-values were generated from regression models comparing the two collection types in each panel. The Bonferroni threshold for this experiment was 0.0125 (4 comparisons) and all statistics which reached significance below this are highlighted in bold. NfL concentrations were natural log-transformed. F Disease group differences between healthy controls and pre-HD in the HD confirmatory cohort from capillary serum samples. p values were generated from multiple linear regressions. Boxes show first and third quartiles, the central band shows the median, and the whiskers show data within 1.5 IQR of the median. The Bonferroni threshold for this experiment was 0.025 (2 comparisons) and statistics which reached significance below this are highlighted in bold. NfL values were natural log-transformed. G Modelling CAG-Age-NfL relationships to show associations between NfL, age, and CAG repeat count using capillary serum samples. Solid lines were produced from our observations using a multiple linear regression model accounting for the main effects of age, CAG, the quadratic effect of age, and the interaction between age and CAG; dashed lines are predictions outside the range of our observations. Separate figures with individual data points for each individual CAG repeat count are provided in Supplementary Fig. 16. HD Huntington’s disease; NfL neurofilament light