Effect of neurofilament analysis on the diagnostic delay in amyotrophic lateral sclerosis

Abstract Aims The aim of this study was to investigate whether neurofilament light (NfL) and phosphorylated neurofilament heavy (pNfH) in cerebrospinal fluid (CSF), sampled prior to referral to a neuromuscular reference center (NMRC), shorten the diagnostic delay in patients with amyotrophic lateral sclerosis. Methods In this retrospective study, patients with ALS were included with (i) determination of neurofilaments (Nfs) before referral to the NMRC (preC‐Nfs ALS, n = 58), (ii) determination of Nfs at the NMRC (C‐Nfs, n = 54) or (iii) with no determination of Nfs (C‐No Nfs, n = 180). Fifty‐six disease controls were included. Results The preC‐Nfs cohort had CSF sampled 2.2 months (range: 0.6–12.0 months) before referral to the NMRC. In this cohort, the diagnostic delay was significantly shorter [median (range): 8.24 (2.37–49.7) months] than in the C‐Nfs cases [median (range): 11.4 (2.93–86.5) months; p < 0.05], but not in the C‐No Nfs cases. When including the disease progression rate and the presence of a genetic mutation as covariates, the difference ceased to exist (p = 0.14). pNfH and NfL levels in the preC‐Nfs cohort were significantly higher than in disease controls (p < 0.0001). Both Nfs showed a similar discriminating performance. Conclusions CSF Nfs assessed before the diagnosis of ALS at a NMRC decreased the diagnostic delay in specific cases by 3 months and only when other covariates were not taken into account.


| INTRODUC TI ON
Amyotrophic lateral sclerosis (ALS) is a progressive disease characterized by the degeneration of both upper and lower motor neurons. 1 Approximately 10% of patients inherit the disease, while others are classified as sporadic ALS. 2 To date, ALS remains a fatal disease as no cure is available. The median survival after symptom onset is 20 to 48 months. 3 Available therapies like riluzole and edaravone have only a modest effect on the disease progression and survival. [4][5][6] Notwithstanding the advances in understanding the disease pathophysiology, the development of therapies significantly affecting the survival in patients with ALS is limited. A contributing factor is the diagnostic delay, which is the time from symptom onset until diagnosis at a neuromuscular reference center (NMRC), or a tertiary ALS referral center. 7,8 In most clinics worldwide, the median diagnostic delay is about 10-12 months. [8][9][10][11][12][13] This delay, which accounts for a significant period of the total survival in patients with ALS, impedes to include patients in clinical trials in the early stage of the disease.
This stage, however, is believed to be the most opportune window to therapeutically slow down the disease progression.
These observations highlight the importance of biomarkers for an early and correct diagnosis of ALS. Fluid biomarkers can be useful tools for clinicians in the diagnostic work-up of patients with ALS.
Neurofilaments have shown great potential as biomarkers in many neurodegenerative diseases including ALS. 14,15 These intermediate class IV filaments are essential for neuronal growth and structural stability of large, myelinated axons. 16 We and others have shown that the neurofilament light chain (NfL) and phosphorylated neurofilament heavy chain (pNfH) are significantly higher in the cerebrospinal fluid (CSF) of patients with ALS, as assessed at a NMRC, than in disease mimics. [17][18][19] Early evidence now demonstrates that CSF NfL levels are elevated 1 year before symptom onset in patients who are genetically at risk for ALS. 20,21 In addition, we showed that pNfH levels in serum are increased in almost 60% of patients with ALS before diagnosis at a NMRC. 22 In this study, we now explore as primary objective if neurofilaments in CSF sampled before referral for diagnosis to a NMRC affect the diagnostic delay in ALS, and as a secondary objective whether early stage determined neurofilaments have a similar diagnostic performance as compared to CSF sampled at a NMRC.

| MATERIAL S AND ME THODS
In total, 292 patients with ALS, all diagnosed at the NMRC of the University Hospitals Leuven according to the revised El Escorial and Awaji criteria, were included between March 2017 and June 2021 in this observational retrospective study. 23 Patients with ALS were divided into three cohorts: (i) 58 patients with ALS who had CSF sampled by a neurologist at a regional hospital to have neurofilament assessed before referral to the NMRC (preC-Nfs), (ii) 54 patients with ALS who had CSF sampled at the NMRC for neurofilament assessment (C-Nfs), and (iii) 180 patients who were diagnosed at the NMRC without any neurofilament assessment (C-No Nfs). The revised ALS functional rating scale (ALSFRS-R) score was obtained at time of diagnosis for all three ALS cohorts. The disease duration was defined as time from symptom onset until sampling. Gene testing for C9orf72, SOD1, TARDBP, and FUS was proposed for patients with ALS using TP-PCR (C9orf72) or direct sequencing (all other genes).
A disease control cohort of 56 patients was consecutively included between February 2017 and December 2020 by local neurologists at regional hospitals. Neurofilaments were requested by the local neurologist for diagnostic purposes, mainly when there was a suspicion of a motor neuron disease. The final diagnoses of the control cohort were obtained through review of the patient's electronic medical records by a neurologist (PM) affiliated to the NMRC of the University Hospitals Leuven (Table S1). Five control patients were excluded as no definite diagnosis was given to the patient.

| The effect of early disease stage determined neurofilaments on the diagnostic delay
Patient demographics of the three ALS cohorts and disease control cohort are presented in Table 1. No significant differences in age or gender were observed among all groups. A concomitant diagnosis of frontotemporal dementia (FTD) was present in 5 (8.6%), 4 (7.4%), and 15 (8.3%) patients in the preC-Nfs, the C-Nfs, and the C-No Nfs cohort, respectively. In addition, phenotypical variants like the flail arm, upper or lower motor neuron predominance, and axial variant were equally present in the three ALS cohorts (Table 1). At time of diagnosis at the NMRC, no significant difference in the distribution of the diagnostic certainty was observed between the ALS cohorts ( Table 1). The median ALSFRS-R score at time of diagnosis and the presence of bulbar onset were not different between the three ALS cohorts (Table 1). A genetic mutation was detected in 5 (10%), 10 (19%), and 25 (19%) patients of the preC-Nfs, the C-Nfs, and the C-No Nfs cohort, respectively (Table S2). In a subanalysis, riluzole therapy was initiated at a later disease stage in the C-Nfs cohort as compared to the preC-Nfs and C-No Nfs cohort (χ 2 = 28.8, Table 1).
The disease duration at sampling in the preC-Nfs cohort (median:   3.2 | Performance of CSF neurofilaments sampled at early disease stage versus sampled at the NMRC Median levels of both pNfH and NfL in CSF were significantly higher in the preC-Nfs cohort and the C-Nfs cohort than in the disease control cohort (p < 0.0001; Figure 3A,B, Table 1). No significant differences were observed in pNfH and NfL levels between the preC-Nfs cohort and the C-No Nfs cohort ( Figure 3A,B, Table 1 Figure 4A,B).
The AUC of CSF pNfH and NfL was 0.84 (95% CI: 0.75-0.90) and 0.85 (95% CI: 0.77-0.91), respectively, to discriminate C-Nfs from disease controls, and were both not significantly different from the AUC of CSF pNfH (p = 0.79) and NfL (p = 0.61) determined between preC-Nfs and disease controls. The sensitivity and specificity of pNfH and NfL levels in CSF to discriminate between C-Nfs and disease controls were similar to those determined between preC-Nfs and disease controls ( Table 2).

| DISCUSS ION
In this study, we now demonstrate that pNfH and NfL in CSF are significantly increased before diagnosis of patients with mainly sporadic ALS. Yet, the availability of CSF neurofilament values in the early stage did not shorten the diagnostic delay of patients with ALS taken into account other covariates.
Interestingly, without taken into account the covariates, the median diagnostic delay was approximately 3 months shorter in patients with ALS whose neurofilament levels were assessed by a local neurologist before referral to the NMRC (preC-Nfs) in comparison with patients whose neurofilament levels were only assessed at time of visiting the NMRC (C-Nfs  delay. Therefore, in a second step, we did implement this covariate in a model to reassess the diagnostic delay between the preC-Nfs and the C-Nfs cohort. We found that the significant difference in diagnostic delay between these cohorts ceased to exist when adding the disease progression rate as a covariate. This indicates that the effect of the disease progression rate on the diagnostic delay might be in imbalance between the two study cohorts. It is known from studies in patients with ALS carrying a genetic mutation that neurofilaments are elevated at an earlier disease stage. Therefore, it might be important to analyze neurofilament levels at an even earlier disease stage, within a prospective study well controlling the disease progression rate in both study arms. 20,21 Nevertheless, it should be noted that the above-discussed covariates are often unknown for a local neurologist at such an early disease stage, but if neurofilaments would be analyzed at such a disease stage these biomarkers can be an incentive to persuade local neurologists to diagnose a patient with ALS or to refer them faster to a NMRC. If patients are referred to a NMRC at an early disease stage, it has been shown to shorten the diagnostic delay with almost 4 months. 25 Moreover, the general practitioner also plays a crucial role to perform additional examinations to quickly refer the patient to a local neurologist and thereby significantly decrease the diagnostic delay in patients with ALS. 26 Other important factors that might have an potential effect on the diagnostic delay, but have not been addressed in this study, are the regional differences of transport, the coverage of regional neurologists, and the social status of patients visiting the hospital with mild symptoms. In line with previous findings from our group and others, we strengthened the evidence that levels of pNfH and NfL are significantly increased in CSF of patients with ALS at time of diagnosis in comparison with disease controls. [17][18][19]27,28 We now explored neurofilament levels in CSF when patients were seen and CSF was sampled by the local neurologist before referral for diagnosis at a NMRC. Interestingly, pNfH and NfL levels were already significantly increased. For CSF NfL, our defined cutoff to discriminate between the preC-Nfs cohort and disease controls was comparable for NfL at the conventional diagnostic stage at a NMRC. For CSF pNfH, the cutoff increased to 1101 pg/ml to discriminate between the preC-Nfs cohort and disease controls, as compared to a cutoff of 750 pg/ ml at the conventional diagnostic stage at a NMRC. 18,19 This stage dependent cutoff for pNfH warrants further investigation. This is, however, in line with our findings that pNfH increases in an early stage to ceil at time of diagnosis at a NMRC. 17 The control cohort included mainly patients with a suspicion of a motor neuron disease, but who in follow-up were not diagnosed with ALS. These patients were often diagnosed with neurological diseases similar to those of disease mimics in previous studies, that is, inclusion body myositis and radiculopathies. [17][18][19]22 Another benefit of the disease controls is the fact that these patients were also sampled in a consecutive order by local neurologists similarly to the patients who later on were diagnosed with ALS, providing an unsupervised matched clinical control cohort. Due to the short follow-up time of the disease controls, a limitation to the study might be that patients are misdiagnosed before obtaining a diagnosis of ALS, as seen by others. 29 The highest specificity toward motor neuron degeneration was observed for CSF pNfH levels, whereas CSF NfL levels yielded the highest sensitivity. This specific pattern fits into the findings that NfL is rather a global early marker of neurodegeneration, while pNfH is more specific to motor neuron impairment as found previously in CSF sampled at a NMRC. 30,31 This is exemplified by a case in our disease control cohort concerning a patient with FTD who had normal levels of pNfH (280 pg/ml) but increased levels of NfL (2835 pg/ml), as seen by others. 30  serve as potential indicator of a motor neuron disease in early disease stages. We already showed that change in serum pNfH levels is much faster in the early stage of ALS than in other early-stage neurological diseases. 21,32 In summary, CSF Nfs assessed before the conventional diagnosis of ALS at a NMRC did not decrease the diagnostic delay when other covariates were taken into account. Given the good discriminating power of Nfs before referral, the added value of Nfs in shortening the diagnostic delay should be further explored prospectively. Onderzoek Flanders, Belgium).

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.