Longitudinal plasma neurofilament light chain and patient-reported outcomes as complementary markers of vincristine-associated peripheral neuropathy in adults with lymphoma: a cohort study Authors: Gretchen A McNally, Grace Ji-eun Shin, Lise Worthen-Chaudhari, Patrick M Schnell, Laura Flora, Surith Sanjay Krishna, Tim Voorhees, Robert Baiocchi, David Bond, Beth Christian, Kami Maddocks, Yazeed Sawalha, Maryam B Lustberg Published: 2026 ABSTRACT AbstractChemotherapy-induced peripheral neuropathy (CIPN) is a common neurotoxicity of cancer treatment with limited diagnostic, monitoring, and treatment options. Neurofilament light chain (NfL) is an axonal cytoskeletal protein released during neuroaxonal injury and a promising biomarker of CIPN, but prospective evidence for NfL as a marker of CIPN from vincristine-containing lymphoma chemotherapy treatment remains limited. To fill this gap, we conducted a pragmatic single-center prospective observational cohort study of adults with non-Hodgkin lymphoma (NHL) receiving first-line vincristine-containing chemotherapy to evaluate NfL dynamics across multiple pre-cycle visits and assess relationships with patient-reported and clinician-graded neuropathy measures. We followed 25 participants during 4–6 months of chemotherapy, and a small subset of those participants (n=6) for 24–42 months post-chemotherapy. Serial plasma NfL was measured and CIPN symptoms were assessed using patient- and clinician-reported measures. Longitudinal changes were analyzed using mixed-effects models. Plasma NfL increased relative to pre-cycle1 at all timepoints (all p<0.001), increasing approximately fourfold by pre-cycle4. Patient-reported CIPN scores and clinician-graded neuropathy also increased during treatment. Exploratory pooled visit-level analyses showed a modest NfL-CIPN association (repeated measures correlation=0.439, p=0.020), while timepoint-specific, lagged, and post hoc sensitivity analyses suggested potential to predict persistent CIPN symptoms from early NfL concentrations. To our knowledge, these findings extend prospective evidence supporting NfL as a biomarker of vincristine exposure in adults with NHL receiving vincristine-containing therapy and may complement patient-reported symptom assessment, clinician grading, and dose-modification context in future CIPN monitoring studies.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s00415-026-14096-2.Keywords: Peripheral neuropathy, Vincristine, Patient-reported outcomes, Neurofilament light chain, Lymphoma, Survivorship Introduction Chemotherapy-induced peripheral neuropathy (CIPN) is a common and clinically consequential adverse effect of vincristine, a core chemotherapy agent within first-line lymphoma treatment regimens [1–3]. Vincristine-associated CIPN has been poorly studied across hematologic cancer populations, and the limited extant literature remains heterogeneous with respect to disease subtype, regimen, neuropathy definition, and assessment method [2–4]. This gap in knowledge regarding CIPN is particularly relevant in non-Hodgkin lymphoma (NHL), a major category of hematologic cancer that includes diverse lymphoid malignancies and is commonly treated with vincristine-containing regimens such as CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) and infusional dose-adjusted EPOCH (etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin), both combined with or without rituximab (R) [1]. The rate at which adults with NHL develop CIPN is high, at up to 75% of patients [3–5]. Estimates vary because of differences in regimen, disease subtype, neuropathy definition, and assessment method [1–5]; however, all studies in this population link neuropathy with sensory symptoms, functional limitations, impaired daily activities, reduced quality of life, and persistent survivorship burden [3, 6, 7]. The broader vincristine-induced CIPN literature describes additional manifestations including neuropathic pain, weakness, impaired proprioception, gait instability, falls, autonomic symptoms, and treatment modification [2, 3, 8, 9]. Mechanistically, vincristine interferes with microtubule assembly and axonal transport, processes essential for maintaining long peripheral axons, thereby linking vincristine exposure to the distal axonal dysfunction and injury that underlie CIPN [10–12]. During neuroaxonal injury, an axonal cytoskeletal protein called neurofilament light chain (NfL) is released and measurable in blood using highly sensitive assays [13–15]. As a blood-based marker of neuroaxonal injury, NfL offers a potentially effective biomarker of CIPN development to complement patient-reported outcomes (PRO) and clinician-graded measures that capture patient experience and symptom severity [16]. In CIPN, blood NfL has been studied most extensively in taxane-, platinum-, and antibody–drug-conjugate-associated neuropathy, with evidence that levels increase during neurotoxic treatment and may correlate with neuropathy severity or persistence in some settings [17–23]. However, evidence supporting the usefulness of NfL as a biomarker of CIPN following NHL chemotherapy is lacking. Challenges to studying NfL as a biomarker for CIPN due to NHL treatment include its lack of specificity. Circulating concentrations can increase across a broad range of central or peripheral nervous system injuries and may also be influenced by non-neuropathy factors such as age, renal function, body size or distribution volume, and neurological comorbidities [13–15, 24–27]. In the CIPN setting, the magnitude, timing, and clinical associations of NfL changes may further vary by anticancer agent, regimen, cancer population, baseline vulnerability, cumulative exposure, and sampling schedule [15, 20, 23]. These challenges in relating NfL dynamics specifically to chemotherapy exposure support the need for longitudinal studies. Ideally, such studies would be pragmatic by aligning repeated measures with clinical visits for vincristine-containing chemotherapy infusions (hereafter, “clinically aligned visits”) in real-world treatment settings, while evaluating NfL alongside established patient-reported and clinician-assessed measures. Another challenge to validating NfL as a biomarker of CIPN is determining how clinicians should interpret a blood-based marker of neuroaxonal injury alongside the patient-reported symptoms and clinician assessments that are currently used to monitor CIPN progression. The timing of patient and clinician perceptions of symptoms can differ from biological changes in ways that are not necessarily linear. Future studies must harmonize NfL dynamics with patient- and clinician-reported outcomes to determine whether NfL provides clinically useful biological context that complements, rather than duplicates, symptom-based assessment. Characterization of NfL dynamics among adults with NHL receiving vincristine-containing regimens is clinically important because survivorship after lymphoma treatment continues to increase, yet neuropathy symptoms persist beyond treatment completion and contribute to long-term symptom burden [3, 6], such that individuals with NHL represent a population in need of improved CIPN care. Therefore, we conducted a pragmatic single-center prospective observational cohort study of adults with NHL receiving first-line vincristine-containing CHOP or EPOCH chemotherapy. We aimed to evaluate the feasibility of clinically aligned serial plasma NfL and neuropathy outcome collection, characterize longitudinal NfL dynamics during treatment, describe concurrent patient-reported and clinician-graded neuropathy outcomes, and explore association with the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire–Chemotherapy Induced Peripheral Neuropathy 20-item instrument (CIPN20), Brief Pain Inventory–Short Form (BPI-SF), National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) neuropathy, dose modification, and longer-term patient-reported outcomes. We hypothesized that plasma NfL would increase with vincristine exposure. By characterizing NfL trajectories alongside symptom, pain, dose modification, and survivorship context, this study addresses an underexplored cancer context and provides a foundation for larger studies testing whether early NfL trajectories can inform CIPN monitoring and long-term symptom risk assessment. Methods Study design This was a pragmatic, single-site, prospective observational cohort study designed to assess the feasibility of the study in the target population and to conduct exploratory analyses to inform future sample size estimates. We performed the study within The Ohio State University (OSU) Arthur G. James Cancer Hospital, a tertiary NCI-designated Comprehensive Cancer Center. The protocol was approved by the OSU Institutional Review Board, and written informed consent was obtained from participants receiving first-line vincristine-containing chemotherapy, CHOP or EPOCH. Study participants and setting Eligible participants were >18 years old, diagnosed with NHL, and initiating first-line vincristine-containing CHOP- or EPOCH-based chemotherapy. The participants were excluded if they had pre-existing neuropathy, central nervous system involvement of lymphoma, non-CIPN neurodegenerative conditions, Eastern Cooperative Oncology Group (ECOG) performance status >2, or concurrent treatment with agents known to be associated with neuropathy (e.g., brentuximab or bortezomib). Disease progression and clinical complications are anticipated in this population and were considered reasons for study discontinuation; this includes participants whose disease progression required discontinuation of the full EPOCH or CHOP protocol or transfer to hospice and/or whose ECOG status increased by >2 while on study. However, participants for whom vincristine was removed, but who otherwise continued on the CHOP or EPOCH regimen, were retained in the study. All study activities occurred within The OSU Arthur G. James Cancer Hospital or within the participant’s home environment. Study timeline Standard treatment regimen includes six chemotherapy cycles administered at 21-day intervals, generally completed within a 4–6-month period. Data collection was aligned with routine clinical care, with research assessments collected during the clinical visits, deferring to the treatment plan regarding when that visit would occur. In addition, among participants who consented to additional cross-sectional follow-up, we collected PROs only at 24–42 months (2–3.5 years), after completion of vincristine exposure for a maximum potential study duration of 4 years. Study activities Data collection platform: data were collected and/or managed using Research Electronic Data Capture (REDCap) tools [28, 29] hosted by OSU. Screening and enrollment: because the study was embedded in routine lymphoma care, enrollment timing depended on clinical workflow and participant availability. Some participants were first approached after treatment initiation and therefore contributed their first research blood and PRO assessments at the next clinically aligned pre-cycle visit, most commonly before cycle 2. We accounted for this by analyzing samples based on actual collection timing rather than treating all first samples as a uniform pretreatment baseline (Fig. 1). Study data collection and analysis schema. A Clinically aligned data-collection schema showing the five labeled timepoints used in the study: T1/Pre-C1, T2/Pre-C2, T3/Pre-C4, T4/Pre-C6, and T5/Post. T1-T4 represent planned treatment-period assessments; T5/Post represents a long-term patient-reported outcome follow-up 24–42 months after chemotherapy completion. B Analysis framework showing how visit-level data contributed to feasibility assessment, the treatment-period plasma NfL model, clinical outcome trajectories, paired exploratory NfL-CIPN20 analyses, and dose-modification/T5 context Medical history at enrollment and throughout study participation: research staff extracted electronic medical record data, including height, weight, medications, comorbidities at diagnosis, clinician-reviewed outcomes, laboratory data, where available, and vincristine dose adjustments at enrollment and throughout study participation. Dose administered per cycle and dose modifications were recorded and analyzed in supplementary data. Dose modification was treated as a clinical context and a potential confounder of NfL and symptom trajectories, but was not assumed to reflect neuropathy unless a neuropathy-related reason was documented. Repeated measures timepoints within routine clinical care: we collected timepoint 1 (T1) just before first chemotherapy exposure in cycle 1 (Pre-C1). As a protocol-defined fallback sampling window, we collected T2/Pre-C2 just before chemotherapy exposure in cycle 2. We prospectively prioritized T1/Pre-C1 collection while allowing T2/Pre-C2 collection as a fallback early-treatment window when T1/Pre-C1 could not be obtained. We collected T3/Pre-C4 and T4/Pre-C6 data just before the fourth and sixth cycles of chemotherapy exposure, respectively. Finally, for a subset of participants, we collected only PROs within a window of 24–42 months after completion of vincristine treatment at T5/Post. T5/Post data were analyzed descriptively and separately from the primary NfL analysis (Fig. 1A). Data collected per timepoint Blood-based biomarker outcomes (primary): peripheral blood was collected from participants at each study timepoint into K3EDTA tubes. Samples were processed by the Hematology Tissue Bank at The Ohio State University within 24 h of collection. Plasma was generated by centrifugation at ambient temperature at 2500 rpm for 10 min. Plasma was aliquoted into 1-mL vials, with two 1-mL plasma aliquots created when sufficient volume was available, and stored at −80 °C until needed. Frozen plasma aliquots were shipped on dry ice to Quanterix (Billerica, MA) for NfL measurement. Plasma NfL was assayed using the ultrasensitive single-molecule array Simoa NF-Light® assay (Quanterix, Billerica, MA). The limit of detection (LOD), lower limit of quantification (LLOQ), and upper limit of quantification (ULOQ) were 0.038 pg/mL, 0.174 pg/mL, and 382 pg/mL, respectively. Quality controls were included in the array and tested with participant samples. Plasma samples and controls were diluted at a 1:4 ratio and measured in duplicate with calibrators. The NfL measurements were performed by Quanterix in two analytical batches. Patient-reported outcomes (secondary): the initial PRO assessment was completed with assistance from a research assistant, whereas subsequent PRO assessments were completed remotely through a REDCap survey link sent to participants by email. These measures capture patient experience of neuropathy and have demonstrated validity and responsiveness in clinical studies [30–34]. Measures collected were as follows: European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-Chemotherapy Induced Peripheral Neuropathy 20-item instrument (EORTC QLQ-CIPN20) [30–32]: The CIPN20 questionnaire was administered as the CIPN-specific PRO instrument representing sensory, motor, and autonomic neuropathy symptoms. Because the driving item and the male sexual-function item in the questionnaire were not applicable to all participants, these items were treated as not applicable when appropriate and were not imputed. We analyzed the raw applicable item CIPN20 score, calculated as the sum of completed applicable items scored from 1 (not at all) to 4 (very much). As a result, the possible minimum raw score varied across participants according to item applicability, ranging from 18 to 20, with higher scores indicating greater patient-reported symptom burden. European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-Chemotherapy Induced Peripheral Neuropathy 20-item instrument (EORTC QLQ-CIPN20) [30–32]: The CIPN20 questionnaire was administered as the CIPN-specific PRO instrument representing sensory, motor, and autonomic neuropathy symptoms. Because the driving item and the male sexual-function item in the questionnaire were not applicable to all participants, these items were treated as not applicable when appropriate and were not imputed. We analyzed the raw applicable item CIPN20 score, calculated as the sum of completed applicable items scored from 1 (not at all) to 4 (very much). As a result, the possible minimum raw score varied across participants according to item applicability, ranging from 18 to 20, with higher scores indicating greater patient-reported symptom burden. Pain severity was assessed using the Brief Pain Inventory-Short Form (BPI-SF) pain severity score, calculated as the mean of four items assessing pain intensity over the past 24 h: worst pain, least pain, average pain, and current pain. Each item is rated on a 0–10 numeric rating scale, with 0 = no pain and 10 = pain as bad as the participant can imagine. The resulting severity score ranges from 0 to 10, with higher scores indicating greater pain intensity [33, 34]. Pain severity was assessed using the Brief Pain Inventory-Short Form (BPI-SF) pain severity score, calculated as the mean of four items assessing pain intensity over the past 24 h: worst pain, least pain, average pain, and current pain. Each item is rated on a 0–10 numeric rating scale, with 0 = no pain and 10 = pain as bad as the participant can imagine. The resulting severity score ranges from 0 to 10, with higher scores indicating greater pain intensity [33, 34]. Clinician-assessed measures (secondary): clinician-assessed measures were obtained from the patient chart into the REDCap database by research staff. The outcome recorded was as follows: National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE v5.0): As part of standard oncology care, treating clinicians performed routine toxicity assessments using CTCAE [35]. The study did not administer CTCAE questions directly; rather, research staff collected neuropathy-related toxicity assessment data as recorded by clinicians during routine care on a five-point scale: 0 (asymptomatic), 1 (mild), 2 (moderate), 3 (severe), 4 (life-threatening). Higher CTCAE grades indicate greater symptom severity and/or functional limitation. Because CTCAE toxicity assessment in routine practice is subject to provider-level variability, particularly for symptoms such as paresthesia that rely on clinical judgment and patient report during the visit, these data were interpreted as clinician-recorded toxicity context rather than as a standardized neurologic endpoint. Patient-reported outcomes were therefore analyzed as complementary measures of participant-experienced neuropathy symptom burden. National Cancer Institute Common Terminology Criteria for Adverse Events (NCI CTCAE v5.0): As part of standard oncology care, treating clinicians performed routine toxicity assessments using CTCAE [35]. The study did not administer CTCAE questions directly; rather, research staff collected neuropathy-related toxicity assessment data as recorded by clinicians during routine care on a five-point scale: 0 (asymptomatic), 1 (mild), 2 (moderate), 3 (severe), 4 (life-threatening). Higher CTCAE grades indicate greater symptom severity and/or functional limitation. Because CTCAE toxicity assessment in routine practice is subject to provider-level variability, particularly for symptoms such as paresthesia that rely on clinical judgment and patient report during the visit, these data were interpreted as clinician-recorded toxicity context rather than as a standardized neurologic endpoint. Patient-reported outcomes were therefore analyzed as complementary measures of participant-experienced neuropathy symptom burden. Chart abstraction for baseline lymphoma stage and glycemic status Baseline lymphoma stage and glycemic status at the time of diagnosis or treatment initiation were abstracted from the electronic medical record. HbA1c was recorded using the value closest to lymphoma diagnosis or treatment baseline when available. When HbA1c was unavailable, the glucose value closest to the diagnosis was recorded and used to calculate a glucose-derived estimate of HbA1c [36]. While measured HbA1c reflects average glycemia over the preceding 2–3 months, HbA1c values estimated from contemporaneous glucose measurements may be disproportionately influenced by short-term hyperglycemia, including glucocorticoid-associated hyperglycemia around chemotherapy initiation [37, 38]. Analysis Sample size determination Enrollment of 25 eligible participants was intended to assess feasibility of clinically aligned serial NfL and patient-reported outcome collection and to generate preliminary estimates for future powered studies. Exploratory biomarker-outcome analyses were interpreted descriptively. Feasibility assessment Feasibility was assessed using prespecified criteria intended to inform the design of a larger longitudinal study. First, collection of an early-treatment assessment was considered feasible if blood and clinical outcome data were obtained from more than 75% of enrolled participants either at T1/Pre-C1 or, when pretreatment collection was not possible because of clinical workflow or later recruitment, at T2/Pre-C2. This criterion was selected to reflect the practical challenges of embedding biospecimen and patient-reported outcome collection into routine NHL care. Second, the rate of retention of participants who remained eligible for study inclusion at T4/Pre-C6 was calculated, and prospective assessment of this final on-treatment timepoint was considered feasible if data collection was achieved among at least 50% of eligible enrollees [39]. This criterion was designed to (a) inform estimates of enrollee attrition due to NHL disease progression between T1 and T4, meaning disease worsening that caused disenrollment from the study due to life-saving updates to the treatment plan, and (b) assess attrition rate among enrollees who remained eligible for study participation. Third, completion of longitudinal assessments was considered feasible if planned treatment-period assessments were collected within the anticipated chemotherapy observation window, allowing for clinically necessary rescheduling of visits due to illness, treatment delays, hospitalization, or other factors. This criterion was designed to evaluate whether clinically aligned longitudinal sampling could be implemented within routine NHL care and to inform allowable timepoint collection windows for future larger studies. We assessed feasibility across these three criteria, requiring that all criteria be deemed feasible to declare the study feasible. Statistical analysis Demographic and clinical characteristics were summarized using descriptive statistics. Continuous variables were summarized using means, medians, standard deviations, minima, and maxima; categorical variables were summarized using frequencies and percentages. Participant codes shown in figures and supplementary tables are randomized public analysis codes generated for this report and do not correspond to study IDs or clinical identifiers. Because plasma NfL concentrations were right-skewed, NfL was analyzed on the log10 scale. The primary longitudinal analysis used a linear mixed-effects model with log10-transformed NfL as the outcome, timepoint as a fixed effect, and participant-level random intercepts to account for repeated measures. Analytical sample sizes differed by observed data availability, and no imputation was performed for any analysis. Model estimates were back-transformed and presented as model-estimated geometric means (pg/mL) with 95% confidence intervals. Relative changes from T1/Pre-C1 were calculated from the model-estimated geometric means. An exploratory regimen model included chemotherapy regimen and timepoint-by-regimen interaction to compare CHOP and EPOCH treatment groups. CIPN20 was analyzed using a linear mixed-effects model with log-10 transformed raw CIPN20 total score as the outcome, timepoint as a fixed effect, and participant-level random intercepts. The estimates were back-transformed for display, and model-based estimated marginal mean contrasts compared each later timepoint with T1/Pre-C1. BPI-SF severity scores and CTCAE neuropathy grades were summarized descriptively and evaluated using fixed effects for timepoint; bootstrap confidence intervals and p values were used because model residuals did not meet the assumptions of normality. T5/Post patient-reported outcome data were analyzed as exploratory long-term follow-up data and displayed separately from the treatment-period visit sequence. T5/Post CTCAE data were descriptive only because only one CTCAE observation was available. Exploratory associations between plasma NfL and CIPN20 scores were evaluated using repeated-measures and Spearman rank correlations because of the small sample sizes and the descriptive, nonparametric nature of these analyses. For the pooled visit-level analysis, all paired treatment-period observations with both plasma NfL and CIPN20 data were included and analyzed using repeated-measures correlation [40]. Exploratory lagged analyses evaluated whether early within-person change in log10 plasma NfL from the first available early-treatment sample to T3/Pre-C4 was associated with subsequent change in CIPN20 total score from T3/Pre-C4 to T4/Pre-C6. Post hoc sensitivity analyses evaluated early T2/Pre-C2 log10 plasma NfL in relation to later CIPN20 total score at T3/Pre-C4 and T4/Pre-C6. Squared Spearman rank correlations (ρ2) were displayed descriptively in selected panels to summarize rank-based monotonic association. Least-squares regression lines were overlaid as visual reference. Exploratory vincristine dose-modification analyses compared participants with no vincristine dose modification versus any vincristine dose modification by T4/Pre-C6. Changes in log10 plasma NfL and CIPN20 total score from baseline to T4/Pre-C6 were compared between groups using Mann–Whitney U tests. All analyses were interpreted in the context of the study's feasibility-focused design and limited sample size. Results Cohort and baseline characteristics Of the 28 individuals who underwent screening, 25 consented and enrolled, as depicted in Fig. 2A per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [41]. The baseline descriptive cohort included 25 participants, of whom 16 were treated with CHOP and 9 with EPOCH (Fig. 2A; Table 1). Baseline demographic and clinical characteristics, including age, sex, race/ethnicity, ECOG performance, lymphoma stage, and glycemic status, are summarized in Table 1. Sixteen were treated with CHOP, which was administered in the outpatient setting, typically during an approximately 1.5-h infusion visit. Nine were treated with EPOCH, which typically required inpatient admission for approximately 5 days and included continuous infusion over 96 h. Participants receiving CHOP were, on average, older than those receiving EPOCH (mean 70.1 vs 53.0 years), more likely to be male (n = 9, 56.3%), and more likely to be White (n = 14, 87.5%). Stage was documented for 18 of 25 participants; among staged participants, 16 had stage III/IV disease and 2 had stage II/IIA disease, while 7 were unstaged or not documented (ND) in the extracted record. Measured HbA1c was available for 19 participants, and glucose-derived estimated HbA1c was available for 6 additional participants. The combined measured/estimated HbA1c median was 5.5% (range, 4.2–8.5); 6 participants had values in the diabetes range (≥ 6.5%, all from CHOP). No participants had concurrent autoimmune disease or were taking a CYP3A4 inhibitor, a potential interactor with vincristine [42, 43], at baseline. One participant with human immunodeficiency virus (HIV) was included in the primary analysis because our previous retrospective study showed no association between HIV and CIPN [44]. Separately, one participant was identified as a high symptom-severity outlier (i.e., ≥2 standard deviations above the mean), and we therefore also report a post hoc sensitivity analysis in which this outlier was removed. Overview of study execution and feasibility. A STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) cohort flow from screening through analytic subsets. Twenty-eight patients were screened; 25 eligible participants contributed to the baseline descriptive cohort. The primary treatment-period plasma NfL model included 24 participants with at least one valid treatment-period plasma NfL measurement, contributing to 52 NfL observations, and the long-term T5/Post PRO subset included 6 participants. B Feasibility criteria and observed results. The early assessment criterion was met (25/25 participants had any assessment; 23/25 had early plasma NfL). Retention through T4/Pre-C6 was met (15/15 had any assessment; 14/15 had PRO data; 12/15 had NfL data). The clinically aligned treatment-window criterion was met, with planned visits completed through T4/Pre-C6 within 6 months Table 1Baseline demographic and clinical characteristics by treatment regimenCharacteristicCHOP (n = 16)EPOCH (n = 9)Total sample (n = 25)Age, mean (SD); range70.1 (13.4); 31–8753.0 (21.2); 19–7463.9(18.4); 19–87Sex, n (%)Male9 (56.3)3 (33.3)12 (48.0)Female7 (43.8)6 (66.7)13 (52.0)Race/ethnicity, n (%)White14 (87.5)6 (66.7)20 (80.0)Non-White2 (12.5)3 (33.3)5 (20.0)Black2 (12.5)1 (11.1)3 (12.0)Asian0 (0.0)0 (0.0)0 (0.0)Hispanic0 (0.0)1 (11.1)1 (4.0)More than 1 race/ethnicity0 (0.0)1 (11.1)1 (4.0)ECOG performance status, n (%)09 (56.3)3 (33.3)12 (48.0)14 (25.0)2 (22.2)6 (24.0)23 (18.8)1 (11.1)4 (16.0)Missing0 (0.0)3 (33.3)3 (12.0)BMI, median (range)29.0 (19.8–45.6)25.9 (20.0–38.5)27.5 (19.8–45.6)Diabetes, yes, n (%)5 (31.3)1 (11.1)6 (24.0)Depression medication, yes, n (%)5 (31.3)0 (0.0)5 (20.0)Cancer stage at baseline, n (%)Stage II/IIA0 (0.0)2 (22.2)2 (8.0)Stage III/IV13 (81.3)3 (33.3)16 (64.0)Unstaged/not documented3 (18.8)4 (44.4)7 (28.0)HbA1c/eA1c, median (range), %6.1 (4.7–8.5)5.4 (4.2–6.0)5.5 (4.2–8.5)HbA1c/eA1c source, n (%)Measured HbA1c12 (75.0)7 (77.8)19 (76.0)Glucose-derived eA1c4 (25.0)2 (22.2)6 (24.0)HbA1c/eA1c category, n (%)Normal (< 5.7%)6 (37.5)8 (88.9)14 (56.0)Prediabetes range (5.7–6.4%)4 (25.0)1 (11.1)5 (20.0)Diabetes range (≥ 6.5%)6 (37.5)0 (0.0)6 (24.0)Baseline demographic and clinical characteristics of the eligible descriptive cohort, stratified by chemotherapy regimen. Continuous variables are summarized as mean (SD) and range or median (range), as indicated. Categorical variables are summarized as n (%). Cancer stage reflects chart-documented stage; unstaged/not documented (ND) indicates no stage information in the extracted record. HbA1c/eA1c combines measured HbA1c with glucose-derived estimated HbA1c when HbA1c was unavailable; glucose-derived eA1c values were used descriptively onlyCHOP cyclophosphamide, doxorubicin, vincristine, and prednisone, EPOCH etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin, ECOG Eastern Cooperative Oncology Group, BMI body mass index, HbA1c hemoglobin A1c, eA1c estimated HbA1c Baseline demographic and clinical characteristics by treatment regimen Baseline demographic and clinical characteristics of the eligible descriptive cohort, stratified by chemotherapy regimen. Continuous variables are summarized as mean (SD) and range or median (range), as indicated. Categorical variables are summarized as n (%). Cancer stage reflects chart-documented stage; unstaged/not documented (ND) indicates no stage information in the extracted record. HbA1c/eA1c combines measured HbA1c with glucose-derived estimated HbA1c when HbA1c was unavailable; glucose-derived eA1c values were used descriptively only CHOP cyclophosphamide, doxorubicin, vincristine, and prednisone, EPOCH etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin, ECOG Eastern Cooperative Oncology Group, BMI body mass index, HbA1c hemoglobin A1c, eA1c estimated HbA1c Feasibility criteria were met All prespecified feasibility criteria were met (Fig. 2B). Early-treatment data collection exceeded the prespecified threshold, with 25 of 25 participants (100%) contributing patient and clinician assessment results at either T1/Pre-C1 or T2/Pre-C2 and 23 of 25 (92%) contributing early NfL data. Retention through the final planned on-treatment assessment, T4/Pre-C6, was achieved for 15 of 15 (100%) for any assessments, while T4/Pre-C6 NfL collection was achieved for 12 of 15 participants (80%). Planned on-treatment assessments were collected at clinically aligned chemotherapy visits through T4/Pre-C6, corresponding to an observation window of up to 6 months. These findings support the operational feasibility of clinically aligned longitudinal sampling in routine lymphoma care, while also identifying NfL-specific retention at later treatment timepoints as an important consideration for future studies. Plasma NfL increased early and remained elevated during treatment Raw NfL distributions were skewed, supporting log10-transformed modeling. In the primary longitudinal mixed-effects model accounting for repeated measures within participants, the model-estimated plasma NfL increased from 26.3 pg/mL at T1/Pre-C1 (95% CI 18.9–36.8) to 91.6 pg/mL at T2/Pre-C2 (95% CI 59.5–141.2), 107.2 pg/mL at T3/Pre-C4 (95% CI 77.3–148.7), and 94.3 pg/mL at T4/Pre-C6 (95% CI 66.1–134.4). Relative to T1/Pre-C1, this corresponded to increases of 248.0% at Pre-C2, 307.4% at T3/Pre-C4, and 258.2% at T4/Pre-C6 (all p < 0.001) (Fig. 3A, B; Fig. S1; Table 2). This pattern is consistent with an early treatment-associated increase in plasma NfL followed by sustained elevation through later on-treatment timepoints, with the highest estimated geometric mean observed at T3/Pre-C4 and persistent elevation at T4/Pre-C6. Because some participants first entered the study at T2/Pre-C2, and T1/Pre-C1 and T2/Pre-C2 values were not available from the same complete set of participants, this contrast should be interpreted as an early collection-timing comparison in the context of heterogeneous baseline timing, rather than uniform within-person pretreatment-to-postexposure change. Longitudinal plasma NfL during vincristine-containing chemotherapy. A Individual participant plasma NfL trajectories shown on the log10 scale across treatment-period visits. Lines connect repeated observations within participants. B Model-estimated plasma NfL geometric means and 95% confidence intervals, back-transformed to pg/mL from the primary mixed-effects model fit on log10-transformed NfL. The model included timepoint as a fixed effect and a participant-level random intercept. Points show back-transformed model estimates and the shaded/error interval shows the 95% confidence interval. P values shown in the panel are model-based estimated marginal mean contrast p values comparing each follow-up timepoint with T1/Pre-C1. Compared with T1/Pre-C1, plasma NfL was higher at T2/Pre-C2 (p = 4.16e−06) T3/Pre-C4 (p = 6.41e−10) and T4/Pre-C6 (p = 3.39e−08). NfL neurofilament light chain Table 2Model-estimated plasma NfL geometric means from the primary mixed-effects modelTimepointEstimated NfL (pg/mL)Lower 95% CIUpper 95% CIT1/Pre-C126.318.936.8T2/Pre-C291.659.5141.2T3/Pre-C4107.277.3148.7T4/Pre-C694.366.1134.4Plasma NfL was analyzed on the log10 scale using a mixed-effects model with timepoint as a fixed effect and participant-level random intercept. Estimates and 95% confidence intervals were back-transformed and are presented as geometric mean plasma NfL concentrations in pg/mLNfL neurofilament light chain, CI confidence interval Model-estimated plasma NfL geometric means from the primary mixed-effects model Plasma NfL was analyzed on the log10 scale using a mixed-effects model with timepoint as a fixed effect and participant-level random intercept. Estimates and 95% confidence intervals were back-transformed and are presented as geometric mean plasma NfL concentrations in pg/mL NfL neurofilament light chain, CI confidence interval A secondary model including the chemotherapy regimen and a timepoint-by-regimen interaction did not identify significant differences between CHOP and EPOCH in this pilot study. The main effect for EPOCH was not significant compared to CHOP (p = 0.779), and timepoint-by-regimen interaction terms were not significant at T2/Pre-C2 (p = 0.132), T3/Pre-C4 (p = 0.929), or T4/Pre-C6 (p = 0.356) compared to T1/Pre-C1 (Fig. S2). Patient-reported and clinical neuropathy measures increased during chemotherapy CIPN20 scores increased with chemotherapy exposure. In the mixed-effects model, model-estimated geometric mean CIPN20 was 20.9 at T1/Pre-C1 (95% CI 19.1–22.8), 20.8 at T2/Pre-C2 (95% CI 18.3–23.6), 23.2 at T3/Pre-C4 (95% CI 21.4–25.1), 25.8 at T4/Pre-C6 (95% CI 23.4–28.4), and 24.5 at >2 years post chemotherapy completion (Post; 95% CI 21.2–28.4). Relative to T1/Pre-C1, this corresponded to a 23.4% increase at T4/Pre-C6 (95% CI 9.1–39.7; p = 0.001). Changes were smaller at T3/Pre-C4 (10.9% increase; 95% CI −0.7 to 23.8; p = 0.070) and Post (17.5% increase; 95% CI −0.2 to 38.4; p = 0.053), while T2/Pre-C2 was essentially unchanged from T1/Pre-C1 (− 0.6%; p = 0.900) (Fig. 4A; Table 3). Patient-reported and clinician-graded outcomes over time. A Model-estimated EORTC QLQ-CIPN20 total score (retained raw calculated total) from the marginal timepoint model output. B Model-estimated BPI-SF severity score from the marginal timepoint model output. C Model-estimated CTCAE neuropathy grade from the marginal timepoint model output. The confidence limits below 0 were truncated at 0 for display. A–C shaded bands show treatment-period estimated marginal means and 95% confidence intervals from T1/Pre-C1 through exploratory T5/Post estimated means and 95% confidence intervals where estimable. Model-based estimated marginal mean contrasts comparing each later timepoint with T1/Pre-C1, shown as p values. T5/Post CTCAE was descriptive only because only one observation was available. CIPN20 Chemotherapy-Induced Peripheral Neuropathy 20-item questionnaire, BPI-SF Brief Pain Inventory-Short Form, CTCAE Common Terminology Criteria for Adverse Events Table 3Model-estimated patient-reported and clinician-graded outcomes by timepointTimepointCIPN20 geometric mean (95% CI)CTCAE mean (95% CI)BPI-SF pain severity mean (95% CI)T1/Pre-C120.9 (19.1–22.8)0.1 (0.0–0.2)0.6 (0.1–1.2)T2/Pre-C220.8 (18.3–23.6)0.3 (0.0–0.9)0.4 (0.0–1.1)T3/Pre-C423.2 (21.4–25.1)0.4 (0.1–0.7)1.4 (0.5–2.5)T4/Pre-C625.8 (23.4–28.4)0.4 (0.1–0.7)1.3 (0.4–2.3)T5/Post24.5 (21.2–28.4)1.0 (n = 1; no CI)2.0 (0.8–3.3)CIPN20 values represent model-estimated back-transformed raw CIPN20 total scores. CTCAE neuropathy grade and BPI-SF severity scores are presented as descriptive model-estimated means with 95% confidence intervals. The T5/Post CTCAE value reflects one available observation and was not used for inferential analysisCIPN20 Chemotherapy-Induced Peripheral Neuropathy questionnaire, CTCAE Common Terminology Criteria for Adverse Events, BPI-SF Brief Pain Inventory-Short Form, CI confidence interval Model-estimated patient-reported and clinician-graded outcomes by timepoint CIPN20 values represent model-estimated back-transformed raw CIPN20 total scores. CTCAE neuropathy grade and BPI-SF severity scores are presented as descriptive model-estimated means with 95% confidence intervals. The T5/Post CTCAE value reflects one available observation and was not used for inferential analysis CIPN20 Chemotherapy-Induced Peripheral Neuropathy questionnaire, CTCAE Common Terminology Criteria for Adverse Events, BPI-SF Brief Pain Inventory-Short Form, CI confidence interval BPI severity scores and clinician-recorded CTCAE toxicity assessment provided additional clinical context and were interpreted descriptively given sparse counts, routine-care ascertainment, and limited sample size. BPI severity scores were generally low during treatment, with mean scores ranging from 0.43 to 1.42 across pre-cycle treatment visits and 2.0 among participants with T5/Post follow-up data (Fig. 4B; Table 3). CTCAE-defined grade ≥1 neuropathy was present in 5 of 15 participants (33.3%) at T4/Pre-C6, including grade 1 neuropathy in four participants and grade 2 neuropathy in one participant. Raw mean CTCAE grade was 0.07 at T1/Pre-C1 (n = 15), 0.29 at T2/Pre-C2 (n = 7), 0.40 at T3/Pre-C4 (n = 20), 0.40 at T4/Pre-C6 (n = 15), and 1.0 at T5/Post (n = 1). In the exploratory CTCAE timepoint model, estimated mean CTCAE grade was 0.1 at T1/Pre-C1 (95% CI 0.0–0.2), 0.3 at T2/Pre-C2 (95% CI 0.0–0.9), 0.4 at T3/Pre-C4 (95% CI 0.1–0.7), and 0.4 at T4/Pre-C6 (95% CI 0.1–0.7). Absolute increases from T1/Pre-C1 were 0.2 at T2/Pre-C2 (95% CI −0.1 to 0.8; p = 0.600), 0.3 at T3/Pre-C4 (95% CI 0.1–0.6; p = 0.005), and 0.3 at T4/Pre-C6 (95% CI 0.0–0.7; p = 0.050) (Fig. 4C; Table 3). Exploratory NfL-CIPN20 analyses suggested group-level association but limited individual-level concordance Because both plasma NfL and neuropathy outcomes increased over treatment, we next explored whether higher NfL was associated with greater patient-reported symptom burden at the visit level or whether earlier NfL change predicted subsequent CIPN20 worsening. In an exploratory repeated-measures correlation analysis combining all paired NfL and CIPN20 observations across treatment timepoints, higher log10 NfL was modestly associated with higher CIPN20 scores (n = 51; repeated-measures correlation = 0.439, p = 0.020). Timepoint-specific relationships were inconsistent and limited by small sample sizes (Fig. 5A; Table S1), and exploratory lagged within-person analyses did not show that early NfL change predicted subsequent worsening of CIPN20. Early change in log10 NfL from the first available early-treatment sample to T3/Pre-C4 was not associated with later change in CIPN20 from T3/Pre-C4 to T4/Pre-C6 (n = 10; Spearman ρ = 0.04, p = 0.91) (Fig. 5B). Exploratory relationships between plasma NfL and patient-reported neuropathy, including sensitivity analyses. A In an exploratory repeated-measures correlation analysis of all paired NfL and CIPN20 observations across treatment timepoints, higher log10 NfL was modestly associated with higher CIPN20 scores (n = 51; repeated-measures correlation r = 0.439, 95% CI 0.079 to 0.698, p = 0.020). Points show paired treatment-period observations colored by visit; grey lines show participant-specific repeated-measures fit lines with a common slope. B Exploratory lagged within-person analysis comparing change in log10 plasma NfL from the first available early-treatment sample to T3/Pre-C4 with subsequent change in CIPN20 total score from T3/Pre-C4 to T4/Pre-C6. Points are colored by chemotherapy regimen; zero reference lines indicate no change. Association was tested using a two-sided Spearman rank correlation: ρ = 0.04, p = 0.91, n = 10. C, D Post hoc sensitivity analyses examining early T2/Pre-C2 log10 plasma NfL in relation to later CIPN20 total score at C T3/Pre-C4 and D T4/Pre-C6. The high symptom-severity outlier (≥ 2 standard deviations above the mean) is highlighted in red but not labeled. Black lines show least-squares linear fits after excluding the highlighted outlier and are included only as visual references. In-panel statistics show two-sided Spearman rank correlations after excluding the highlighted outlier: C ρ = 0.441, p = 0.381, n = 6; D ρ = 0.900, p = 0.037, n = 5. Displayed ρ2 values represent squared Spearman correlations. With the highlighted outlier included, the corresponding Spearman correlations were C ρ = −0.109, p = 0.816, n = 7, and D ρ = 0.086, p = 0.872, n = 6. These post hoc analyses are exploratory and underpowered and should be interpreted as evidence of individual-level heterogeneity rather than validation of a predictive NfL-CIPN20 relationship. NfL neurofilament light chain, CIPN20 Chemotherapy-Induced Peripheral Neuropathy questionnaire, CHOP cyclophosphamide, doxorubicin, vincristine, and prednisone, EPOCH etoposide, prednisone, vincristine, cyclophosphamide, and doxorubicin A post hoc sensitivity analysis further examined the relationship between early NfL and later CIPN20. We excluded one data point determined to be an outlier, because of a symptom severity score more than 2 standard deviations outside the mean and a biologically justified reason for exclusion (i.e., additional disease burden associated with immunodeficiency), the association between T2/Pre-C2 NfL and T4/Pre-C6 CIPN20 became positive by rank correlation, but the analysis remained based on very small numbers and should be interpreted only as hypothesis-generating. This sensitivity analysis supports the presence of individual-level heterogeneity and reinforces that plasma NfL and CIPN20 should not be treated as interchangeable measures in this pilot dataset (Fig. 5C, D). Dose modifications occurred for a quarter of participants Vincristine dose modifications—including reductions, holds, and discontinuations—occurred for 6 out of 25 (24%) of the cohort with reasons that included neuropathy, ileus, constipation, and multifactorial clinical judgment (Table S2). No effect of dose modification was detected (Fig. S3A, S3B). As depicted in Fig. S3A, however, dose-modified participants had a numerically larger median increase in log10 NfL from baseline than participants without dose modification by T4/Pre-C6 (median Δlog10 NfL, 0.632 vs 0.244; approximately 4.3-fold vs 1.75-fold increase; Mann–Whitney p = 0.194). In contrast, median CIPN20 change from baseline to T4/Pre-C6 was lower in the dose-modified group than in the no-dose-modification group (median ΔCIPN20, 0 vs 6; Mann–Whitney p = 0.138). These exploratory comparisons did not meet conventional statistical significance and were limited by small sample size, heterogeneous timing, and heterogeneous reasons for dose modification. However, they suggest that the delivery of vincristine and dose modification may alter the observed relationship between NfL kinetics and patient-reported symptom trajectories. Exploratory long-term patient-reported outcomes suggested CIPN persistence Long-term patient-reported outcome data were available for six survivors 24–42 months after completion of vincristine-based chemotherapy. At T5/Post, raw CIPN20 total scores ranged from 20 to 36, and BPI-SF worst-pain item ranged from 0 to 4.25. The two participants with the highest T5/Post CIPN20 scores also had the highest T5/Post BPI severity scores as well as vincristine dose modification during treatment. It is important to note, however, dose modifications cannot be assumed to have occurred because of neuropathy worsening. Indeed, dose modifications did not lead universally to high T5/Post symptom burden (Table S3). Complete treatment-period NfL values from the first available early-treatment sample through T3/Pre-C4 and T4/Pre-C6 were available for four participants. Given the small number of participants with complete NfL trajectories and long-term PRO data, these long-term data are presented descriptively to show that patient-reported neuropathy symptoms can persist years after therapy and that real-world dose modification is an important interpretive variable when evaluating NfL and symptom trajectories. Further studies are required to test the hypothesis that a specific NfL trajectory shape, dose reduction, and functional outcome measures can serve as a multidomain risk model for persistent neuropathy. Discussion Feasibility criteria were met, and our hypothesis that plasma NfL would increase with vincristine exposure was upheld. We were able to collect serial plasma NfL within the course of routine care, administer neuropathy (CIPN20) and pain (BPI) patient-reported symptom assessments at the point of care or via emailed surveys, and abstracted clinician-recorded CTCAE neuropathy grades, vincristine dose-modification information, and relevant baseline information from the medical chart review. The resulting cohort data demonstrate that plasma NfL increased after initial vincristine exposure and remained elevated across later chemotherapy infusion cycles. Exploratory repeated measures analyses of the relationship between NfL and CIPN revealed that the measures increased together when timepoints were pooled; however, at individual timepoints, NfL and CIPN20 outcomes showed no correlation within this small dataset. The post-hoc sensitivity analysis suggested that early NfL changes may predict later CIPN20 changes. These findings support further study of plasma NfL as a potential biomarker of CIPN and related symptom burden resulting from vincristine-containing lymphoma chemotherapy. Mechanistically, our interpretation that plasma NfL concentrations are sensitive to vincristine exposure is plausible. A preclinical rodent study indicates increased NfL following vincristine treatment, further supporting the biological plausibility of a vincristine-associated NfL response consistent with vincristine-associated distal axonal injury [22]. In addition, recent Blood/ASH abstract reports, including Gotti et al. in pediatric acute lymphoblastic leukemia and our contemporaneous report in adults with lymphoma treated with CHOP or dose-adjusted EPOCH, provided prospective human evidence that serum/plasma NfL can be measured longitudinally in vincristine-associated neuropathy contexts [44, 45]. The present study extends these prior observations, which were published as conference abstracts but not full manuscripts, by characterizing longitudinal plasma NfL dynamics alongside patient-reported and clinician-graded neuropathy outcomes in adults with NHL receiving vincristine-containing therapy. Our results showed an overall increase in NfL, but this increase did not follow a simple monotonic cumulative-dose pattern typically reported for other chemotherapeutic agents such as platinum and paclitaxel [15, 17–21, 23]. Our data are in line with a long-term follow-up study of vincristine-treated participants with hematological cancer, as well as with early-onset vincristine-induced peripheral neuropathy in B-cell lymphoma, which showed no association with the total vincristine dose or the number of treatment cycles [3, 4]. Together, these findings support the notion of a potentially distinct biology underlying vincristine-induced NfL release and neurodegeneration contributing to CIPN. Emerging mechanistic data suggest that NfL release may also intersect with injury-associated immune responses. A recent study reported that NfL released after neuronal damage can activate myeloid cells and promote neuroinflammatory responses in a model of amyotrophic lateral sclerosis [46]. This may be relevant to vincristine-associated neurotoxicity because vincristine is increasingly understood to involve immune-mediated processes in preclinical studies, including neuroimmune signaling, astrocyte activation, macrophage/myeloid responses, cytokine release, and pain sensitization [47–53]. Although our study was not designed to test whether NfL contributes functionally to neuroinflammation in CIPN, these observations motivate investigating plasma NfL in the context of a broader injury and immune response rather than as a same-day symptom-severity readout. Immune modulation of vincristine-induced neurotoxic injury may help explain why NfL and CIPN20 were related at the pooled-visit level but showed limited concordance at the individual level. Patient-perceived symptoms may emerge before, during, or after detectable NfL release, reflecting sensory nerve dysfunction, altered excitability, immune activation, pain sensitization, and functional adaptation. Conversely, circulating NfL may remain elevated after the peak of acute symptom worsening if neuroaxonal injury, clearance kinetics, or neuroimmune responses persist. The post hoc sensitivity analysis further supports the possibility of individual-level heterogeneity. However, this analysis was underpowered and hypothesis-generating; it should not be interpreted as evidence of a predictive NfL-CIPN20 relationship. Rather, it reinforces the more conservative conclusion that NfL and CIPN20 are complementary but non-interchangeable measures. Clinician-recorded CTCAE neuropathy and patient-reported CIPN20 should not be expected to align perfectly because routine-care CTCAE grading reflects clinician-documented toxicity during clinical encounters [35], whereas CIPN20 captures symptom burden through a structured patient-reported instrument [30, 31]. In this cohort, CIPN20 showed a clearer longitudinal increase in patient-reported symptoms than clinician-graded CTCAE neuropathy scores, consistent with prior studies indicating that clinician- and patient-reported CIPN measures are related but not interchangeable [54–57]. Similar temporal divergence between clinician-assessed neuropathy and patient-reported burden has also been reported in the NOVIT study of hematologic patients receiving vincristine-, bortezomib-, or lenalidomide-containing therapy [58]. Consistent with this broader literature, Andersen et al. (2024) emphasized that NfL reflects neuroaxonal injury and should be interpreted alongside clinical and patient-reported outcomes rather than as a direct readout of symptom severity [15]. Together, these observations support multi-modal assessment and reinforce that NfL, clinician grading, and PROs provide complementary rather than redundant information. The BPI severity provided additional clinical context but should not be interpreted as a CIPN-specific endpoint, as it captures overall pain intensity from cancer, treatment, comorbidities, musculoskeletal conditions, or other causes. In this cohort, BPI severity scores were generally low during treatment, whereas CIPN20 scores increased. This pattern is consistent with the idea that patient-reported neuropathy burden may include numbness, tingling, sensory change, motor symptoms, and functional impact that are not fully captured by a general pain severity measure. Several limitations should be emphasized. First, this was a small, single-center pilot study with substantial missingness for PROs, clinician-recorded CTCAE toxicity grading, BPI severity score, dose-modification variables, and complete longitudinal NfL trajectories. Second, early sampling was heterogeneous because clinically aligned collection in routine lymphoma care did not always permit T1/Pre-C1 sampling before first vincristine exposure; therefore, T2/Pre-C2 served as a protocol-defined fallback early-treatment window. This limits interpretation of early timepoint contrasts and reinforces the need for standardized pretreatment sampling in future studies. Third, NfL is not specific to CIPN and can be influenced by age, renal function, body size/distribution volume, neurological comorbidity, and assay factors [13–15, 24–27]. These factors may contribute to between-participant variability, particularly in a small cohort, and support interpreting NfL longitudinally rather than as a cross-sectional diagnostic threshold. Fourth, clinician assessment was limited to standard-of-care CTCAE toxicity documentation. The study did not administer CTCAE questions directly, and toxicity grading in routine practice is subject to provider-level variability, particularly for symptoms such as paresthesia that rely on clinical judgment and patient report during the visit. The study also did not include standardized neurologic examination, nerve conduction studies, quantitative sensory testing, or tissue-level validation. This limits the ability to distinguish patient-reported symptom burden, clinician-recorded toxicity grade, and objective neurologic impairment. Fifth, dose modifications were not available in a fully adjudicated, subject-level, time-varying dataset, limiting formal analysis of delivered vincristine exposure, dose reduction, and dose omission as dynamic clinical variables. In addition, dose changes can occur for reasons other than development of CIPN, so such modifications cannot be assumed to be neuropathy-related in every case; they are best interpreted as real-world treatment modifications that may alter subsequent vincristine exposure, NfL release, and symptom trajectories. Finally, T5/Post data were exploratory, PRO-based, and available only in a small subset; they should not be used to estimate the prevalence of chronic CIPN or to assess predictive performance. Despite these limitations, this study provides an important prospective reference point for interpreting plasma NfL in adults receiving vincristine-containing lymphoma therapy. The cohort captures a clinically realistic treatment setting, with pre-cycle sampling, patient-reported symptom assessment, routine-care CTCAE toxicity grading, BPI severity assessment, dose-modification context, and exploratory long-term PRO follow-up. The robust early increase in plasma NfL supports its sensitivity to vincristine-containing treatment exposure and neuroaxonal injury biology, while the incomplete alignment among NfL, CIPN20, CTCAE toxicity, BPI severity, and dose modification demonstrates why NfL should not be interpreted as a stand-alone symptom or toxicity measure. In this respect, the study is valuable not because it establishes a predictive biomarker, but because it defines the key methodological and interpretive issues that larger vincristine-focused CIPN biomarker studies must address. Future studies should use standardized pretreatment sampling, larger longitudinal cohorts, post-treatment and survivorship follow-up, prespecified lagged analyses, subject-level dose-intensity data, and clinician-confirmed neurologic endpoints. Larger cohort studies incorporating clinical assessments are needed before NfL can be used to inform routine CIPN clinical decisions [16]. Conclusion Plasma NfL increased substantially during first-line vincristine-containing CHOP- or EPOCH-based chemotherapy in adults with NHL, with an early rise and sustained elevation through later chemotherapy exposure. Patient-reported neuropathy symptom burden also increased during treatment. However, NfL did not serve as a direct surrogate for patient-reported symptom severity, clinician-recorded toxicity, pain severity, or dose-modification context at any single timepoint. Exploratory analysis revealed a possible association between early NfL changes and later CIPN20 severity. Larger studies are needed to determine whether longitudinal NfL trajectories can help predict CIPN symptom burden and inform intervention strategies for adults with NHL. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 358 KB) Acknowledgements We are grateful to the patients who participated in this study. We thank the Hematology Tissue Bank Shared Resource and the Recruitment, Intervention and Survey Shared Resource at The Ohio State University Comprehensive Cancer Center for research blood processing, REDcap database development and study support. Finally, we would like to thank Dr. Bhuvana Ramaswamy for her invaluable support. Author contributions G.A.M. designed the study, coordinated the work, interpreted the data, wrote the initial manuscript, and revised the manuscript. G.J.S. contributed to conceptual framing, interpretation of results, statistical analyses, and writing and revision of the manuscript. L.W.-C. contributed to conceptual framing, statistical interpretation, and writing and revision of the manuscript. P.M.S. performed and advised on statistical analyses. L.F. and S.S.K. contributed to data collection. T.V., R.B., D.B., B.C., K.M., and Y.S. contributed to patient care and clinical interpretation. M.B.L. contributed to study conception, clinical interpretation, and manuscript review. All authors read and approved the final manuscript. Funding This work was supported by The Ohio State University 2020 Intramural Supportive Care Pilot Research Award Program (G.A.M.), NIH R01-AG084676 (L.W.-C., M.B.L., and G.A.M.), NIH K01AR080833 (G.J.S.), and NIH R21CA274588 (G.J.S.). Data availability The datasets generated and/or analyzed during the current study are not publicly available due to institutional and IRB requirements for participant-level clinical and biomarker data, but deidentified data may be available from the corresponding author on reasonable request and subject to institutional and IRB approval/requirements. Code availability Analysis code may be available from the corresponding author upon reasonable request and subject to institutional and IRB requirements. Clinical trial registration: Not applicable. This was an observational cohort study and participants were not assigned to an intervention. Declarations Conflicts of interest The authors have no relevant financial or non-financial interests to disclose. Ethical approval This study was performed in accordance with the principles of the Declaration of Helsinki. The study protocol was approved by The Ohio State University Institutional Review Board. Informed consent Written informed consent was obtained from all individual participants included in the study. Consent for publication Not applicable. The manuscript does not contain identifying individual participant information or identifiable images. Participant codes shown in figures and supplementary tables are randomized public analysis codes and do not correspond to study IDs or clinical identifiers. Footnotes Gretchen A. McNally and Grace Ji-eun Shin have contributed equally to this work. References Associated Data Supplementary Materials Supplementary file1 (DOCX 358 KB) Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available due to institutional and IRB requirements for participant-level clinical and biomarker data, but deidentified data may be available from the corresponding author on reasonable request and subject to institutional and IRB approval/requirements. Analysis code may be available from the corresponding author upon reasonable request and subject to institutional and IRB requirements. Clinical trial registration: Not applicable. This was an observational cohort study and participants were not assigned to an intervention.