Longitudinal plasma phosphorylated neurofilament-heavy chain levels in oxaliplatin-induced peripheral neuropathy: a prospective multi-institutional exploratory study

Abstract

Purpose

Reliable circulating biomarkers for chemotherapy-induced peripheral neuropathy (CIPN) remain lacking. Phosphorylated neurofilament-heavy chain (pNF-H) is an axon-oriented biomarker of neuroaxonal injury, but its clinical relevance in oxaliplatin-induced peripheral neuropathy has not been established. We evaluated the feasibility and longitudinal behavior of plasma pNF-H in patients receiving oxaliplatin-based chemotherapy.

Methods

This prospective multi-institutional exploratory study included patients with unresectable or recurrent colorectal cancer receiving CapeOX plus bevacizumab as part of the NYX study, a randomised trial of ninjin’yoeito (NYT; TJ-108, Tsumura & Co., Tokyo, Japan), a twelve-ingredient Kampo herbal extract, at 9.0 g/day for 24 weeks in the NYT arm. Plasma samples were collected at baseline, at clinically assessed CIPN onset (follow-up 1), and at the time of neuropathy resolution, at cycle 9, or at discontinuation, whichever came first (follow-up 2). Plasma pNF-H concentrations were measured by enzyme-linked immunosorbent assay. CIPN was assessed using CTCAE version 5.0.

Results

Thirty patients with evaluable plasma samples were included. Plasma pNF-H levels showed substantial inter-individual variability without a consistent temporal pattern. Compared with baseline and CIPN onset, plasma pNF-H levels at follow-up 2 were modestly but significantly increased. However, plasma pNF-H levels were not associated with the onset, severity, or persistence of clinically assessed CIPN.

Conclusions

Although longitudinal plasma pNF-H measurement was technically feasible, pNF-H did not function as a clinically meaningful biomarker for oxaliplatin-induced peripheral neuropathy. These findings may reflect a pathophysiological mismatch between dorsal root ganglion neuronal cell body injury and the axon-oriented biological specificity of pNF-H.

Introduction

Oxaliplatin-based chemotherapy is widely used for unresectable or recurrent colorectal cancer; however, chemotherapy-induced peripheral neuropathy (CIPN) remains one of its most frequent and clinically significant toxicities, often leading to dose reduction or treatment discontinuation and long-term impairment of quality of life [1–3]. Despite its clinical importance, objective biomarkers capable of reflecting neuroaxonal injury in CIPN are lacking.

Current assessment of CIPN relies mainly on patient-reported symptoms and clinician-based grading systems such as the Common Terminology Criteria for Adverse Events (CTCAE). These approaches are inherently subjective and insufficiently sensitive to subclinical nerve damage [2, 3]. Accordingly, there is an unmet need for biologically meaningful circulating biomarkers.

Neurofilaments are neuron-specific cytoskeletal proteins released into the circulation following axonal injury. Neurofilament light chain has been extensively studied as a biomarker of neuroaxonal damage in various neurological disorders and has recently been explored in the context of CIPN [4–6].

In contrast, phosphorylated neurofilament-heavy chain (pNF-H), which reflects injury to large-caliber myelinated axons, has been less well investigated in oncology. However, oxaliplatin induces loss of pNF-H neuronal immunoreactivity in rat dorsal root ganglion (DRG) tissue [7]. In non-CIPN neurological conditions, pNF-H has been reported to be associated with disease severity or clinical outcome, including acute intracerebral hemorrhage [8] and amyotrophic lateral sclerosis [9], supporting its biological relevance as a marker of neuroaxonal injury. Whether these findings can be extrapolated to CIPN remains uncertain.

Oxaliplatin-induced peripheral neuropathy is increasingly recognized as a form of neuronopathy, in which dysfunction of dorsal root ganglion neuronal cell bodies precedes secondary axonal degeneration, raising questions about the suitability of axon-oriented circulating biomarkers for disease monitoring.

Within the framework of the NYX study in patients receiving CapeOX plus bevacizumab therapy, plasma pNF-H was incorporated as an exploratory biomarker [10]. The present study aimed to evaluate plasma pNF-H as an exploratory translational neurotoxicity biomarker in oxaliplatin-induced peripheral neuropathy.

Materials and methods

This prospective, multi-institutional exploratory biomarker study was conducted between 2019 and 2023 as part of the NYX study, a randomised trial of ninjin’yoeito (NYT; TJ-108, Tsumura & Co., Tokyo, Japan; a standardised Kampo herbal extract of twelve botanical ingredients) administered at 9.0 g/day for 24 weeks versus CapeOX plus bevacizumab alone. The pNF-H sub-study was conducted in parallel in consenting patients from both arms; full protocol details are published elsewhere [10]. Patients with unresectable or recurrent colorectal cancer receiving CapeOX plus bevacizumab were eligible for inclusion; eligibility and exclusion criteria are detailed in the protocol paper [10]. Botanical and pharmaceutical specifications of NYT are reported according to STORK (Standards of Reporting Kampo products) [11].

Plasma samples were collected at three predefined time points: at baseline before initiation of chemotherapy, at the time of clinically assessed CIPN onset during treatment (follow-up 1), and at the time of neuropathy resolution, at cycle 9, or at discontinuation, whichever came first (follow-up 2). Blood was collected into EDTA-containing vacuum tubes (Venoject II, VP-DK052K, Terumo, Tokyo, Japan) and centrifuged at 1,900×g for 15 min at 4 °C. Plasma aliquots were stored at − 20 °C or below until analysis. The 500 mg/m² cumulative oxaliplatin threshold for follow-up 1 sampling was selected based on published evidence for the median onset of CIPN grade 2–3 at this dose [12–14].

Plasma pNF-H concentrations were determined by enzyme-linked immunosorbent assay using a commercially available human pNF-H ELISA kit (BioVendor, Laboratory Medicine, Brno, Czech Republic) according to the manufacturer’s protocol. The analytical measurement range was 62.5–4000 pg/mL, with a limit of detection and quantification of 23.5 pg/mL. The intra-assay coefficient of variation was less than 4.5%. Values below the limit of quantification were estimated using the calibration curve whenever possible. All measurements were performed in duplicate in a blinded manner.

CIPN was evaluated using CTCAE version 5.0 at each time point.

Comparisons of plasma pNF-H levels between time points were performed using paired t-tests. Between-group comparisons of plasma pNF-H levels by CTCAE grade were performed using the Mann–Whitney U test. Associations between plasma pNF-H levels and CTCAE grade were evaluated using Spearman rank correlation analysis. A two-sided P value of < 0.05 was considered statistically significant.

Ethics approval and informed consent: this study was approved by the institutional ethics committee of Kanazawa Medical University (approval number CRB4180007) and participating institutions and was conducted in accordance with the ethical standards of the institutional research committees and with the 1964 Declaration of Helsinki and its later amendments. Written informed consent was obtained from all participants prior to enrollment.

Results

A total of 78 patients were enrolled in the NYX study, of whom 30 were included in the final analysis due to the availability of evaluable plasma samples (Fig. 1).

Fig. 1.

Study flow diagram. Of 78 patients enrolled in the NYX study, 52 had plasma samples submitted for pNF-H measurement, and 30 were included in the protocol-defined interim analysis. Among these, 20 had samples available at all three time points (baseline, follow-up 1, and follow-up 2), 7 at baseline and follow-up 1 only, and 3 at baseline and follow-up 2 only

Longitudinal trajectories of plasma pNF-H demonstrated marked inter-individual variability without a consistent temporal pattern (Fig. 2).

Fig. 2.

Longitudinal plasma pNF-H trajectories in individual patients. Open circles represent patients treated with ninjin’yoeito (NYT; a traditional Japanese Kampo medicine) plus CapeOX and bevacizumab; closed circles represent patients treated without NYT (non-NYT). Lines connect values from the same individual. Baseline, follow-up 1 (at clinically assessed CIPN onset), and follow-up 2 (at the time of neuropathy resolution, at cycle 9, or at discontinuation, whichever came first) are shown

Compared with baseline and follow-up 1, plasma pNF-H levels at follow-up 2 were modestly but significantly higher (baseline vs. follow-up 2: mean difference + 68.2 pg/mL, P = 0.019; follow-up 1 vs. follow-up 2: mean difference + 99.9 pg/mL, P = 0.0034). Although plasma pNF-H levels at follow-up 2 were modestly but significantly higher than those at baseline or at CIPN onset, these longitudinal changes were not associated with the onset, severity, or persistence of clinically assessed CIPN. Spearman rank correlation analysis revealed no significant association between plasma pNF-H levels and CTCAE grade at either follow-up 1 (ρ = 0.12, P = 0.56, n = 24) or follow-up 2 (ρ = −0.02, P = 0.93, n = 19). At follow-up 1, median plasma pNF-H levels were 40.8 pg/mL (IQR 13.1–59.2, range 2.3–69.2) for Grade 1 (n = 13), 48.5 pg/mL (IQR 16.9–78.7, range 4.6–138.5) for Grade 2 (n = 10), and 22.3 pg/mL (n = 1) for Grade 3. At follow-up 2, median plasma pNF-H levels were 121.9 pg/mL (IQR 21.7–349.2, range 3.9–656.1) for Grade 1 (n = 14) and 95.4 pg/mL (IQR 76.2–160.8, range 39.2–276.1) for Grade 2 (n = 5), with no significant difference between groups (Mann–Whitney U test, P = 0.96). The wide IQR observed within each grade category, particularly for Grade 1 at follow-up 2, reflects the substantial inter-individual variability in plasma pNF-H levels irrespective of CIPN severity.

Discussion

In this prospective multi-institutional exploratory study, longitudinal plasma pNF-H dynamics were evaluated; however, pNF-H levels did not reflect clinically meaningful chemotherapy-induced peripheral neuropathy.

Although statistically significant longitudinal increases in pNF-H were observed after completion of oxaliplatin-based chemotherapy, these changes were modest and characterized by substantial inter-individual variability. Importantly, plasma pNF-H levels showed no correlation with the severity or persistence of CIPN as assessed by CTCAE, indicating that pNF-H does not function as a clinically useful surrogate marker in this setting. Therefore, the clinical relevance of pNF-H in oxaliplatin-induced CIPN should be interpreted with caution. Formal correlation analysis confirmed no significant association between plasma pNF-H levels and CTCAE grade at either time point.

Previous studies have demonstrated that pNF-H is associated with disease severity or clinical outcome in several neurological disorders, including acute intracerebral hemorrhage and amyotrophic lateral sclerosis [8, 9]. The present findings suggest that such associations may not translate directly to oxaliplatin-induced peripheral neuropathy, reflecting differences in pathophysiology and patterns of axonal injury. To our knowledge, this is the first prospective study to evaluate longitudinal plasma pNF-H in oxaliplatin-induced peripheral neuropathy.

Recent studies have reported associations between neurofilament light chain (NfL) and CIPN severity [4–6]. Preclinical studies from the NeuroDeRisk consortium have further demonstrated that plasma NfL — in contrast to pNF-H — is a sensitive biomarker of peripheral neurotoxicity in oxaliplatin- and paclitaxel-treated rodents, with NfL concentrations rising proportionally with nerve lesion grade and NfL emerging as the most reliable blood-based marker across multiple neurotoxic exposures [15–17]. In contrast, pNF-H primarily reflects structural axonal degeneration. The discrepant findings between NfL and pNF-H may therefore reflect differences in the neuroaxonal components captured by each biomarker. Additionally, background CNS-derived neurofilament signals may obscure any CIPN-specific pNF-H response given the limited biomass of DRG neurons relative to CNS neurons. Furthermore, bevacizumab-mediated VEGF-A suppression may have attenuated symptomatic CIPN expression in our cohort independently of axonal injury [15, 18], contributing to the dissociation between structural damage and clinical grade.

Exploratory subgroup analyses suggested a tendency toward higher pNF-H levels in patients with more severe chronic CIPN, although these findings were not statistically significant and should be interpreted cautiously given the limited sample size.

No significant correlation with pNF-H was observed at follow-up 1 (ρ = 0.23, P = 0.32, n = 21) or follow-up 2 (ρ = −0.23, P = 0.40, n = 16), confirming that pNF-H does not reflect CIPN severity regardless of neuropathy phenotype [12–14]. This neuronal cell body dysfunction is thought to precede and subsequently lead to secondary axonal degeneration, rather than primary distal axonal injury. In contrast, taxane-induced neuropathy is generally considered to involve primary axonal injury related to microtubule dysfunction and impaired axonal transport, with neuronal cell body involvement occurring secondarily in some cases.

Preclinical studies support DRG neuronal injury as the key pathological substrate of chronic oxaliplatin-induced neuropathy [19].

Given that pNF-H predominantly reflects structural axonal injury, the limited association between circulating pNF-H levels and clinical CIPN severity observed in the present study may be explained by this disease model. In this context, pNF-H elevation may represent a downstream consequence of cumulative neuronal injury originating at the neuronal cell body, rather than a direct marker of primary disease activity. Accordingly, the present findings should be interpreted not as a lack of biological relevance, but as a reflection of the mismatch between the dominant pathological site in oxaliplatin-induced neuropathy and the biological specificity of pNF-H.

However, several important limitations should be considered when interpreting the present findings. First, the assessment of CIPN relied solely on CTCAE version 5.0 at each time point. Although CTCAE is widely used in oncology trials, it is inherently subjective and may not adequately capture subtle or early neuropathic changes, nor distinguish precisely between acute and chronic neuropathic processes. The lack of patient-reported outcome measures or objective neurophysiological assessments, such as nerve conduction studies, further limits the sensitivity of CIPN evaluation in this study. Other limitations warrant consideration, including the modest sample size and the exploratory nature of the biomarker analysis. Second, the protocol-defined sampling trigger (cumulative oxaliplatin ≥ 500 mg/m²) restricted follow-up sampling to patients who developed clinician-assessed CIPN; patients without CIPN were not sampled, potentially introducing selection bias. Third, the sampling trigger may itself lag behind the underlying neurobiological injury [20]. Fourth, pNF-H was the sole pre-specified biomarker at study design (2019); the absence of a comparative NfL measurement limits interpretation within the current biomarker landscape [4–6, 15–17]. Nevertheless, the prospective multi-institutional design and standardized longitudinal sampling strengthen the validity of the present findings.

In conclusion, longitudinal plasma pNF-H levels showed limited association with clinically assessed CIPN in this exploratory study. These findings suggest that axon-oriented circulating biomarkers may inadequately capture the pathophysiology of oxaliplatin-induced neuropathy and highlight the need for multimodal biomarker strategies incorporating both clinical and neurophysiological assessments.

Author contributions

Y. Mo. and M. M. conceptualized. M. M. performed formal analysis. Y. Mo, K. S., T. T., S. I. Y. Mu., A. T. investigated. M. M. tool responsibility of methodology. : S. I., T. T., N. Y., Y. Mu provided resources. A. T. supervised. M. M. validated the results. Y. Mo. visualized. Y. Mo. wrote original draft, and all authors reviewed the manuscript.

Funding

This study was supported by Tsumura & Company.

Data availability

Data are available from the corresponding author upon reasonable request.

Declarations

Conflict of interest

Y. Mo. has received lecture fees from Tsumura & Company. N. Y. has received research grants and lecture fees from Tsumura & Company. S. I. has received lecture fees from Tsumura & Company. All other authors declare no conflicts of interest.